This action might not be possible to undo. Are you sure you want to continue?

**Mechanical Systems and Signal Processing
**

Mechanical Systems and Signal Processing 20 (2006) 2112–2147

www.elsevier.com/locate/jnlabr/ymssp

**A model-based identiﬁcation method of transverse cracks in rotating shafts suitable for industrial machines
**

` Bachschmid, Andrea Vania Paolo PennacchiÃ, Nicolo

Dipartimento Di Meccanica, Politecnico di Milano, Via La Masa 34, I-20156, Milano, Italy Received 4 February 2005; received in revised form 8 April 2005; accepted 13 April 2005 Available online 13 June 2005

Abstract This paper presents a model-based transverse crack identiﬁcation method suitable for industrial machines. The method is validated by experimental results obtained on a large test rig, which was expressly designed for investigating the dynamical behaviour of cracked horizontal rotors. The identiﬁcation method and the relative theory is brieﬂy presented, while three different types of cracks are considered: the ﬁrst is a slot, therefore not actually a crack since it has not the typical breathing behaviour, the second a small crack (14% of the diameter) and the third a deep crack (47% of the diameter). The excellent accuracy obtained in identifying position and depth of different cracks proves the effectiveness and reliability of the proposed method. r 2005 Elsevier Ltd. All rights reserved.

Keywords: Rotor dynamics; Crack; Identiﬁcation; Crack depth; Crack position

1. Introduction Early papers documenting crack occurrences in real machines date back to the end of 1960s and to 1970s [1–4]. Propagating transverse cracks have been discovered in the last 20 years [5,6] in several rotors of steam turbines or generators of European power plants. Fortunately, as far as

ÃCorresponding author. Tel.: +39 2 2399 8440; fax: +39 2 2399 8492.

E-mail addresses: paolo.pennacchi@polimi.it (P. Pennacchi), nicolo.bachschmid@polimi.it (N. Bachschmid), andrea.vania@polimi.it (A. Vania). 0888-3270/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.ymssp.2005.04.005

ARTICLE IN PRESS

P. Pennacchi et al. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2113

the authors know, they have been detected before the crack had propagated to a critical depth, that means before the occurrence of a catastrophic failure. The importance of early detection of cracks, possibly by means of an automatic diagnostic methodology that uses the information furnished by standard monitoring systems, which generally analyse the vibrations measured in correspondence of the bearings only, appears obvious from these considerations. Many different publications deal with the problem of crack identiﬁcation in rotating shafts [7–11], but the methods seem to be unsuitable for application to real machines in operating conditions and have never been tested nor validated with experimental results. In Imam’s et al. paper [7], the effect of the presence of a crack is emphasised by huge bending loads and the crack detection is validated by means of laboratory tests and measurements, which are inapplicable to real machines during operation. It must be reminded that the possibility of detecting a crack depends highly from the position where the crack has developed, from loads that are applied to the shaft and from the points where the symptoms are measured. Also crack-related thermal effects can help in identifying its presence. Sekhar [9] uses a model-based approach in the time domain, which has been developed by Bach and Markert [8]. Anyhow, the method is tested only using numerical results obtained on very simple rotor models, which are unable to represent the behaviour of real machines. Dilena and Morassi [10] use antiresonances in the frequency response curves of non-rotating beams, while Gounaris and Papadopoulos [11] use the axial vibrations of cracked shafts to detect position and depth of cracks. However, industrial rotating machinery are generally not equipped with sensors able to measure axial vibration. Thus, both methods seem unsuitable to detect cracks in rotating machinery during operation in industrial plants. Other approaches are presented by Dong et al., who use the deﬂections of the ﬁrst mode shape at two symmetric points and the contour diagram of crack location versus crack depth for the ﬁrst two normalised eigenfrequencies [12]. Also, this method can be hardly applied on real machines. Seibold and Weinert [13] use banks of Kalman ﬁlters in the time domain on measured vibrations to identify crack depth and localisation, but the method requires the deﬁnition of large number of banks of ﬁlters depending on the accuracy of crack location and depth identiﬁcation. Another similar application is presented in Bucher and Seibold [14]. The dynamical behaviour of rotors with transverse crack has been studied by many authors (extensive surveys are given in [15–17]) and some typical symptoms of crack are well known in a rotating shaft of a machine equipped with standard sensors: a change in 1Â rev., 2Â rev. and 3Â rev. vibration vector is suspect. A change in vibration vector means not just an increase or a decrease in amplitude, but also a change in phase only with constant amplitude. However, 1Â rev. components can be caused by many other faults (e.g. unbalance, bow, coupling misalignments) and 2Â rev. components can be due also to polar stiffness asymmetries (in generators), to surface geometry errors (journal ovalisation) and to non-linear effects in oil ﬁlm bearings. These two last causes can also generate 3Â rev. components. It is then extremely important to have a reference situation, stored by the monitoring system, in which the behaviour of the rotor system without faults and in similar operating conditions is analysed. The reference situation, better than by a steady-state condition at normal operating speed, is represented by run-down transient which furnishes much more information about its dynamical behaviour. By comparing then the actual behaviour during a run-down transient with

ARTICLE IN PRESS

2114 P. Pennacchi et al. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147

the reference behaviour, the change in vibrations can be evaluated and by means of one of the automatic diagnostic procedures based on fault-symptom matrices or on decision trees approach, the type of the most probable impending fault can be identiﬁed. Once the type of fault has been identiﬁed in a shaft line, also its most probable position and its severity should be identiﬁed. This is then possible by means of the least squares approach in the frequency domain, which is described in the following. The method is validated on a test rig specially designed for analysing the dynamical behaviour of horizontal cracked rotors. Three types of crack are taken into account: the ﬁrst type is a slot of 34% depth, which is not actually a crack and has not the typical breathing behaviour and which excites signiﬁcantly only the 2Â rev. component. Even if this situation does not generally reproduce a developing crack in a real machine, it is in any case interesting since many papers in literature take into consideration a slot, because the generation of a real fatigue crack in a shaft is a difﬁcult (and expensive) task. The second case, a 14% partially breathing crack is much more interesting in real application since it represents a crack in the early stage of its development; the possibility of early detection of smaller cracks increases the safety of the plant and allows to undertake corrective action. The third case, relative to a about half diameter crack (47%) represents a rotor operating in a dangerous situation, in which the rotor condition is rather close to safety threshold. From this point of view it is very interesting to be able to establish, from vibration measurements during normal operation only, the real depth of the crack.

2. Model-based identiﬁcation Using the approach ﬁrstly outlined in [18], the method has been developed and presented in [19]. As described in [20], assuming a ﬁnite beam element model for the rotor, the effect of a crack on the statical and dynamical behaviour of the rotor can be simulated in the frequency domain, by applying to the rotor different sets of equivalent forces in correspondence of the cracked beam element, one set for each one of the three harmonic components. It can be shown that the overall behaviour of an horizontal axis heavy cracked shaft is linear. Only in extreme operating conditions, the non-linear effect of the breathing crack, which is weak in normal conditions, may inﬂuence its behaviour. The problem of the identiﬁcation of the position of the crack is then reduced to an external force identiﬁcation procedure, described in [21]. The stiffness of a cracked shaft is periodical due to the breathing and the rotation of the crack. The Fourier’s expansion of the periodical stiffness is truncated in correspondence of the third harmonic component ½KðOtÞ ¼ ½Km þ ½DK1 eiOt þ ½DK2 ei2Ot þ ½DK3 ei3Ot þ Á Á Á . (1) The elements of ½DKn are different from 0 only in the positions corresponding to the cracked element nodes. Introducing this stiffness in the equations of motion of the rotor, Eq. (2) is obtained ! 3 X _ t þ ½Km þ € t þ ½Cx ½DKj eijOt xt ¼ Fe þ W, (2) ½Mx

j ¼1

the reference case vibrations xref are given by Eq. thus the force system is not in general represented by rotating forces. (5) is ﬁnally obtained. Ã Ã i2Ot 1 1 Ff 2 ei2Ot ¼ Àð½DK2 xs þ 1 . Reðxn einOt Þ ¼ 1 2 xn e 2 xn e (6) Similarly. (9)–(11) are. Eq. between the measured vibration xtotal of a rotor system that has a fault and the reference case xref .) and W is the statical force vector (weight and so on). represents the vibrational behaviour due to the fault. (6) and (7) is limited to the third harmonic component.ARTICLE IN PRESS P. Note that the projections along reference axes of the harmonic component of the force system are not necessarily equal. the ﬁrst. The displacement xt can be split in its statical xs and dynamical x components xt ¼ xs þ x. since they are due to the impending fault only. (5) If the nth harmonic component of the vibration x is taken into account. the second and the third harmonic components of the crack force system. If also the product in the last term of Eq. Ã Ã Ã Ã iOt 1 1 1 Ff 1 eiOt ¼ Àð½DK1 xs þ 1 2½DK2 x1 þ 2½DK3 x2 þ 2½DK1 x2 þ 2½DK2 x3 Þe . (12) _ ref þ ½Kxref ¼ Fe € ref þ ½Cx ½Mx (12) . Pennacchi et al. which is called sometimes additional vibrations. (5) is limited to the third harmonic component. € þ ½Cx _ þ ½Km x ¼ Fe À ½Mx 3 X j ¼1 (3) (4) ½DKj eij Ot ðxs þ xÞ. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2115 where O is the rotating speed. the difference xa ¼ xtotal À xref . ½M is the mass matrix. According to Eq. These vibrations are used in the identiﬁcation procedure. then the following equivalent force component vectors are obtained: Ã Ã Ã Ã Ã Ã 1 1 1 1 1 Ff 0 ¼ Àð1 4½DK1 x1 þ 4½DK1 x1 þ 4½DK2 x2 þ 4½DK2 x2 þ 4½DK3 x3 þ 4½DK3 x3 Þ. the nth component of Fourier’s expansion of the periodical stiffness ½DKn is given by inOt Ã ÀinOt Reð½DKn einOt Þ ¼ 1 þ1 . From the experimental point of view and under the previously exposed hypothesis of linearity. (3). etc. ½C the damping matrix (including gyroscopic effects). where the statical component is given by xs ¼ ½Km À1 W. bow. Fe the external rotating forces (unbalance. In fact. 2½DK3 x1 þ 2½DK1 x1 þ 2½DK1 x3 Þe i3Ot 1 Ff 3 ei3Ot ¼ Àð½DK3 xs þ 1 2½DK1 x2 þ 2½DK2 x1 Þe _ _ _ _ (8) (9) (10) (11) of which Eqs. 2 ½DKn e 2 ½DKn e (7) The Fourier’s expansion of both Eqs. respectively. it can be expressed as inOt Ã ÀinOt þ1 .

For sake of clarity and simplicity. from which it differs only very little: the crack affects the stiffness of one element only. it can be approximated by ½K of the uncracked shaft.yÞ x. the ﬁrst one is often neglected in comparison to the second one. Therefore. Further.ARTICLE IN PRESS 2116 P. the three ﬁrst harmonic components of the vibration are considered. Anyhow. In the right-hand side (r. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 while those caused by the developing crack are given by € total þ ½Cx _ total þ ½Km xtotal ¼ Fe þ Ff 0 þ Ff 1 eiOt þ Ff 2 ei2Ot þ Ff 3 ei3Ot . the following equations are obtained for each harmonic component: ½ÀðnOÞ2 M þ inOC þ KXn ¼ Ff n . also ½Km is unknown. as it has been done in the Bernoulli’s beam in comparison to the Timoshenko’s beam.s. Pennacchi et al.yÞ where fFð which has all null-elements except for the dof to which the L g is the localisation ðvector x. (13) is considered for an unknown crack.yÞ is a complex number representing the amplitude and the phase forcing system is applied.yÞ . (15). 2. (16) x. xÞ ðyÞ The equivalent force system acts on few dofs of the system.) of Eq. In fact. in case of 4 degrees of freedom (dofs) per node model. therefore generally huge bending moments are present in correspondence of cracked elements. Due to energy considerations. the set of equivalent forces in the case of a crack is reduced to a couple of bending moments (with their relative phase) along two orthogonal directions for each nth harmonic component. it can be shown that in a beam element loaded by shear forces and bending moments. a single fault is considered. ½Mx _ _ _ _ (13) If Eq. respectively. In the case considered in the paper. For a crack located in the j th element of . while multiple faults can be handled using the method fully described in [21]. The ﬁnal equations are recalled here below. € a þ ½Cx ½Mx _ _ _ _ (14) By applying the harmonic balance criteria in the frequency domain and considering the additional vibrations Xn . the equivalent force system is applied to the two nodes of the element that contains the crack and is therefore composed by a vector of eight generalised forces. has to be identiﬁed. in which a crack is going to be identiﬁed.yÞ ¼ fFð Fð L gA fn ðx. Therefore. Therefore the additional vibrations are given by _ a þ ½Kxa ¼ Ff 0 þ Ff 1 eiOt þ Ff 2 ei2Ot þ Ff 3 ei3Ot . are not full-element vectors that are convenient to be represented by means of x. n ¼ 1. it results that the most important among these forces are the bending moments which are roughly equal and opposite on the two nodes of the cracked element. For this reason. therefore vectors Fð f n and Ff n . the elastic deﬂection energy associated to the shear forces is much smaller than the energy associated to the bending moments. each one of the reference system directions. and A of the fault. 3 (15) in which the force vector Ff n .h. the additional deﬂections due to the crack can be attributed to equivalent bending moments only. cracks develop and propagate mainly to axial stresses generated by bending moments in rotating shafts.

3. 2. 0 |ﬄﬄﬄ{zﬄﬄﬄ} ¼ ðyÞ yÞ fFð Lj gAn . Since the fault has to be identiﬁed not only in its severity but also in its position. > > .ARTICLE IN PRESS P. by considering that the vibration measuring points in real machines are corresponding to bearing location. introducing the admittance matrix ½EðnXÞ of the system ½EðXÞ ¼ ½ÀðnXÞ2 M þ inXC þ K. . . Pennacchi et al. . (15) becomes 2 EðnO1 Þ 0 . 7 . in general the procedure has to be repeated per each node of the rotor.0 foundation dofs 0 ﬄÁ{zﬄﬄﬄ Á Á ﬄ0 |ﬄﬄﬄ } xÞ ijn gT Á M ð n e ðxÞ n ¼ 1. unless the research of the fault is limited in a speciﬁed interval of the nodes. 0 n ¼ 1. . (20) is overdetermined since the number of the observation (the measured vibrations at different rotating speeds) are greater than the number of the parameters of the fault that have to be identiﬁed. . 0 0 ﬄ0 |ﬄﬄﬄi{zﬄﬄ } ðxÞ j th rotor node j þ1th rotor node xÞ ¼ fFð Lj gAn . . so the additional vibrations are available for several rotating speeds and a set of np rotating speeds is considered X ¼ fO1 O2 Á Á Á Onp gT . xÞ . yÞ . Eq. . . (20) Anyhow. 2. the corresponding equivalent force systems are: . 2. EðnOnp Þ Xn 0 (19) (18) 6 6 6 ½EðnXÞXn ¼ 6 6 4 n ¼ 1. 3. 7> . The fully assembled model dofs are ð4nr þ nf Þ while per each rotating speed are measured only 2nb dofs. . is nr À 1. j th rotor node j þ1th rotor node foundation dofs n ¼ 1. to the rotational horizontal and vertical dofs of the extreme nodes of element j . / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2117 the shaft line. ð17Þ where the only terms different from zero are those relative. . Then. Fð 001 f n ¼ f0 . . respectively. The procedure used to solve the problem is the following. Note that ne . and the supporting structure is represented by means of a foundation with nf dofs. . Eq. . 3. The admittance matrix ½EðnXÞ has order ðð4nr þ nf Þnp Â ð4nr þ nf Þnp Þ. Eq. number of elements. 38 9 Xn > > > > > > 7 > 0 = 7< Xn > 7 ¼ Ff n . 3. (20) represents the general system of equations for all the dofs of the considered fully assembled machine. 0 0 EðnO2 Þ . |ﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄ} 0 ﬄÁ{zﬄﬄﬄ Á Á ﬄ0 |ﬄﬄﬄ } yÞ ijn gT Á M ð n e ðyÞ . (15) can now be rewritten considering that condition monitoring systems collect data for many rotating speeds.0 0 0 0 À1 . 2. 0 0 0 . Fð f n ¼ f0 . . > > 5> > : > . Let the machine model have nr nodes and nb bearings in which the vibrations XBmn are measured at np rotating speeds in two orthogonal directions. the equivalent force fault identiﬁcation problem in Eq. 0 À i 0 ﬄ0 |ﬄﬄﬄﬄﬄ ﬄ{zﬄﬄﬄﬄﬄ } .

(21) in the r. Weighted least squares technique can be employed in order to increase the robustness of the identiﬁcation as described in [22]. FL1 ¼ . starting from the ﬁrst element and moving along the rotor from element to element. are separated from the remaining dofs of the system.e. 2. has to be estimated. the module and the phase for each harmonic component) of the equivalent force systems. n ¼ 1. the vibrations of the dofs. that ﬁts best the experimental data XBmn . This is done by ﬁrst substituting Eq.ð1Þ = ð1Þ n ^ Bn ÞÀ1 ½X ^ Bn T XBmn . of Eq.ð1Þ eijðnxÞ. min k½X n n n ¼ 1. the localisation matrix for the ith rotating speed is 2 3 0 i 0 0 0 Ài 0 0 0 0 0 T 6 7 . 2. the phase and a residual of the fault. which are measured. ðxÞ ðyÞ (24) Now the complex vector An ¼ fAn An gT (i.ð1Þ ijð n : Mð . 6 7 . due to unitary force The effect on the measured dofs ½X n systems applied to the selected element of the model is now calculated. The equivalent force system.ð1Þ yÞ. 3. . (20). . e n (26) .h. ½FL1 ðnp Þ (22) ^ B . which vector is of order ð2nb np Â 1Þ. obtaining the matrix ½HðnXÞ.s. 3 (25) whose general solution is given by means of the pseudo-inverse calculation 8 9 < M ðxÞ. ÁÁÁ ðiÞ 6 7 (21) ½FL1 ¼ 6 7 0 0 1 40 0 0 0 À 1 0 0 0 |ﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄ} 5 1st rotor node 2nd rotor node foundation dofs and for all the np rotating speeds 8 9 ½FL1 ð1Þ > > > > < = . 3. by considering only the rows of ½HðnXÞ corresponding to the measured dofs. (23) Then. applied to the selected element. n ¼ 1. is applied in each element of the rotor model. ½X dofs n ¼ 1. 2. and inverting matrix ½EðnXÞ. Pennacchi et al. The problem is equivalent to ^ B An À XBm k. 2. (24) is obtained: ^ Bn ¼ ½HðnXÞjmeasured FL1 . / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Now. ^ Bn T ½X An ¼ ¼ ð½X yÞ. Therefore. least squares identiﬁcation is used here in order to evaluate the module. so for all the rotating speeds the fault vector is of order ðð4nr þ nf Þnp Â 2Þ. The partitioned matrix is of order ð2nb np Â ð4nr þ nf Þnp Þ and following Eq. The ﬁtting is done in least squares sense. since the number of the unknown (the module and the phase) is less than the equations (recalling that data are corresponding to several rotating speeds and each of the sets is composed by several measuring planes). of order ð2nb np Â 1Þ. > > > > : . in the ﬁrst element. Xn ¼ ½EðnXÞÀ1 Ff n ðXÞ ¼ ½HðnXÞFf n ðXÞ. 3.ARTICLE IN PRESS 2118 P. .

2. If a fault only is considered.ARTICLE IN PRESS P. component is therefore the most suitable symptom for detecting position and depth of the crack. A full discussion would be far from the scope of this paper. described in the previous section. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 ð1Þ 2119 The module and the phase of the complex value in the rows of An are the equivalent bending moments due to the crack. is obtained: 1Þ ðnr À1Þ drn ¼ ðdð Þ. 29 and 37). (28). The identiﬁcation procedure. but not for the identiﬁcation of its depth. The 3Â rev. Crack depth identiﬁcation The following procedure has been implemented for the identiﬁcation of the crack depth. 3. XBn ¼ ½X n n ¼ 1.e. identiﬁes the crack position in a particular element of the rotor. . ordered by the element number. Pennacchi et al. (29) A graph representing the values of drn along the shaft axis (from the ﬁrst to the last beam ﬁnite element) can be drawn. the highest values are obviously reached during a run-down transient when approaching the resonant condition at 1 2 critical speed. drn n ¼ 1. Often this component can be recognised only when approaching the resonant condition at a rotating speed equal to 1 3 the rotor’s critical speed. . 1). a set in R of relative residuals given by Eq. 3. n ¼ 1. 1 shows also the crack depth p. but a reader potentially interested in measures of accuracy in model-based identiﬁcation can refer to [23]. 21. component is useful for the localisation of the crack.) and equal to l (Fig. . . whose length is known from the ﬁnite element model (f. whose estimation is given by the ðyÞ xÞ ijn Þ ðyÞ ijn e þ M e of Eq. . The ﬁnal remark about the model-based method regards the accuracy of the machine model. the better the estimation of the faults is per each harmonic component. component is rather small and generally masked by some noise. (29) is. the 1Â rev. The 2Â rev. in the selected element. components) that are applied to this element. (26). Fig. 3 (27) by calculating the difference with the measured vibrations and then by normalising it !1=2 ÃT ½ X À X ½ X À X Bn Bmn Bn Bmn 1Þ dð . rn ¼ T X XÃ Bmn Bmn (28) The procedure is then iterated for all the ne elements of the rotor. 3. 2Â and 3Â rev. when no other sources of bow are present. It is worth noting that the 1Â rev. These equivalent bending moment components M1 . Therefore. The closer to zero the minimum corresponding values Mn ¼ M ð n n value of Eq. vibration components are due both to the breathing mechanism of the crack and to the local bow which generally has developed during the crack propagation. the residual in the selected element is determined. 2. by ﬁrst obtaining the calculated response due to the identiﬁed fault in the selected element ^ Bn Að1Þ .m. rn . 2. The location that corresponds to the minimum value of drn indicates the most probable location of the faultðx(see Figs. the dofs of the extremity nodes of the element and the bending caused by the equivalent moment components Mn (1Â. Finally.

i. it is possible to show that the projections xÞ ðyÞ along the reference axes are equal in case of 2Â rev. and the equivalent force system is a rotating moment of amplitude M 2 . since the statical deﬂection. is calculated from model data.ARTICLE IN PRESS 2120 P.e. component. Then the statical bending moment M in correspondence of the same element. (17) and (30). especially when operating conditions far from resonances in critical speeds are analysed. it follows 3 2 9 0 i 0 0 0 À i 0 0 T8 < M ðxÞ eijðnxÞ = n 60 0 0 1 0 0 0 À 17 ¼ À½DKn jj xs jj 4 |ﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} 5 yÞ yÞ ijð n . M ð 2 ¼ M 2 ¼ M 2 and ðxÞ ðyÞ j2 ¼ j2 . horizontal axis. Even if vector Ff n has order ð4nr þ nf Þ Â 1. M2 and M3 have been calculated from the corresponding 1Â. Moreover. we obtain Ff n ¼ À½DKn xs . industrial rotating machinery are considered. If we neglect all terms of Eqs. Pennacchi et al. 1) xs jj ¼ fxj Wxj yj Wyj xjþ1 Wxjþ1 yjþ1 Wyjþ1 gT . : Mð n e j th rotor node j þ1th rotor node (32) . (9)–(11). measured vibrational behaviour. The moments M1 . is much higher than the vibrations. it is more convenient to consider only the sub-matrix ½DKn jj of order 8 Â 8 corresponding to those dofs and the displacement vector xs jj of the nodes of the cracked element (Fig. 1. M2 and M3 represent the amplitude of the equivalent crack forces deﬁned. the only elements different from zero are those corresponding to the dofs of the cracked element j th. respectively. (9)–(11) except the ﬁrst one. in the shaft sections loaded by consistent bending moments. by Eqs. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 x j+1 xj ϕn Ω node j ψn Mn j ϑy j+1 node j+1 z ϑx j+1 Mn ψn y j+1 j+1 ϑy j p ϑx j yj l Fig. due to the weight and to bearing alignment conditions. Cracked element. (30) This assumption is reasonable if heavy. 2Â and 3Â rev. |ﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} j th rotor node j þ1th rotor node (31) By recalling Eqs. Therefore.

ﬁxed in space. (34) in Eq. Moment M is then proportional to the relative rotations of the extremity nodes of the cracked element. 2) as shown in [20. since ½DKn jj and ½Km jj of the cracked element are function of p only. therefore only the statical deﬂections of the nodes of the cracked element are required in Eq.e. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2121 and for the modulus M n ¼ kMn k ¼ k½DKn jj xs jj k.24]. Fig. .ARTICLE IN PRESS P. force vector W acting on the cracked element. This equivalent length l c has been tuned by means of numerical 3D analyses (Fig. i. as in the Bernoulli’s beam. the mean value of the stiffness matrix of the cracked element. cjþ1 À cj (see Fig. These bending moments do not change much from node to node of a element and can therefore be considered constant in the cracked element and equal to M . xs jj is composed mainly by rotational deﬂections of the extremity nodes. Since in the crack. 2. (33) Matrix ½DKn is restricted to the nodes of the cracked element. 1). (33). By considering the sub-matrix ½Km jj of order 8 Â 8 corresponding to those dofs extracted from the matrix ½Km . Equivalent cracked beams models used in 3D analyses. (35) shows that ratio M n =M depends only on the crack depth p. Pennacchi et al. |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} j th rotor node j þ1th rotor node (34) By replacing Eq. if regular shapes with rectilinear tips are considered. instead of the length l of the cracked element in the model of the shaft. the statical deﬂections of the cracked element are given also by xs jj ¼ ð½Km jj ÞÀ1 f0 cos js 0 sin js 0 À cos js 0 À sin js gT M . A similar ratio M 0n =M has been calculated for different cracks depths considering the equivalent length l c of the cracked element. Therefore xs jj can be calculated by considering the statical bending moments due to static. M (35) Eq. it ﬁnally results M n ¼ k½DKn jj ð½Km jj ÞÀ1 kM ! Mn ¼ k½DKn jj ð½Km jj ÞÀ1 k. (33). the shear force can be neglected with respect to bending moment ﬂexibility.

Bending moments ratio on the equivalent cracked beam.5 1 0. of the rotor. bending moment component M n applied to one element of the f. component and expressed by the relationship in Eq. Anyhow. (37). as shown in [20]. 2Â and 3Â rev. M 0n ¼ f ðpÞ. as a function of crack relative depth for the nÂ rev. is also depending on the relative crack depth p lc ¼ gðpÞ. EJ (36) This is represented in Fig. (37) can then be used for determining the crack depth. beam element that simulates the behaviour of the cracked beam. It is worth noting that the nÂ rev.5 2 M′/M 1.e. Eq. the equivalent bending moments M n are applied to an element with an actual length l instead of l c . which is proportional to the product M n Á l of the identiﬁed nÂ rev.5 0 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Crack relative depth Fig. However. 4. component only is present and 1Â and 3Â rev. 3. component.m. measured displacements are the effect of the relative rotation of the cracked element extremity nodes. coming from the identiﬁcation procedure.ARTICLE IN PRESS 2122 P. the length l c of the equivalent. M (37) In the same ﬁgure also the curve M 02slot =M for an always open crack (a slot or notch) is shown: in this case the 2Â rev. multiplied by its length l . Pennacchi et al. component are absent. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 3 M1 ′ /M M2 ′ /M M3 ′ /M M2 ′ slot /M 2. 3 for the 1Â. reduced stiffness. D (38) The function gðpÞ is represented in Fig. Moment M 0n gives the same relative rotation of the nodes of the beam with length l c as M n does with a beam with length l Dcn ¼ M 0n l c. .

25 0. The equivalent bending moment component M 0n . Fig.1 0. can therefore be calculated as M 0n Á l c ¼ M n Á l . the diameter D and the length l c of equivalent beam. the same M can be considered applied to the element with . 4. f(p) g(p) 2x rev. applied to an equivalent cracked beam element of length l c .6 0. Function for the calculation of the crack depth. 1.2 1 f(p) g(p) 0.4 1.05 0 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Crack relative depth 2123 Fig. Relationship between the crack relative depth p. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 0. (39) By assuming that the statical bending moment M applied to the original element of length l does not change along the element. 5. Pennacchi et al.2 0.35 0.2 0 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Crack relative depth f(p) g(p)1x rev.45 0.4 f(p) g(p) value 0.4 0.5 0.15 0.ARTICLE IN PRESS P.8 0.6 1. f(p) g(p)slot 2x rev. f(p) g(p) 3x rev.3 lc /D 0.

ARTICLE IN PRESS 2124 P. mainly steam turbines of power plants. (38) we get Mn Á l ¼ f ðpÞ Á gðpÞ.88 m. is composed by a shaft The test rig EUROPE of EDF (Electricite divided in three parts supported by two equal oil ﬁlm bearings. The bearings are three-lobed shape. 2-bearing 1-composed shaft test rig on rigid foundation. (37) we can derive M 0n M n Á l ¼ ¼ f ðpÞ M Á lc M and using Eq. Recalling Eq. The proximity probes for the measurements of relative shaft-journal vibrations are installed very close to the bearings.s. Pennacchi et al. 6. The nominal diameter of the shaft is 70 mm and the overall length is 3. The distance between the bearings is 1. 6. 5 for the nÂ rev. 4.. Experimental results 4. The supporting structure can be considered as rigid in the speed range 0–1500 rpm. the relative depth of the crack. from the known l. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 equivalent length l c . The total mass is 450 kg and the main inertia disk has a mass of 250 kg. components.h.15 m. Test rig assembly ´ de France). 6. (41). shown in Fig. . they are both on the bearing side towards the motor). In this conﬁguration the 1st critical speed is close to 1150 rpm. M ÁD (41) (40) Eq. but not inside of them. can then be used for determining. Fig. The test rig has been designed in order to be as much as possible representative of industrial machines. shown in Fig. as usually occurs in real machines (in Fig.1.

the maximum speed was reduced to 1500 rpm. but a relevant part of this depth is due to the notch. In this application for safety reasons. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bending Moment [Nm] 1000 500 0 -500 -1000 2125 Left side Right side 15 20 23 29 Central part RAFT© 2. Since the central part of the rotor. Fig. 7. 7).1 mm and of relative depth of 34% was generated by electroerosion. 8. From the analysis of Fig. In the ﬁrst case a transverse slot of thickness 0. 8. Mean crack depth of 14% in the second experimental case. at a distance of 310 mm from the main inertia disk. The crack position along the rotor corresponds again to element 20. Pennacchi et al. between elements 15 and 29.0 0 0. 7 and is composed 45 ﬁnite beam elements. the crack average depth results of only about 14% with a maximum depth of 11. equipped with 4 poles generators. At the end of the tests this last shaft was broken to check the actual shape and depth of the crack.3 mm equal to 16. Finite beam element model of the test rig and statical bending moment distribution. . The last is also the operating speed of turbo-groups of many nuclear power plants. assembled in the rotor system. and used for validating the identiﬁcation procedure.5 1m Fig. a crack was obtained as a consequence of fatigue solicitation. can be disassembled. three types of crack were generated on three different specimens. The model of the rotor is shown in Fig. This position corresponds roughly to element 20 in the ﬁnite beam element model of the rotor (Fig.1%. In the second case. starting from a notch made by electroerosion. Maximum design speed is 3000 rpm.ARTICLE IN PRESS P.

The statical bending moments are shown in Fig. component in Figs. checked at the end of the tests. while the dashed line the moments on the r. This bow is generally different from that present in the reference case. the cracked part presents usually a permanent bow due to the fatigue solicitation used to generate the crack.s. due to the reason previously expressed. The possibility of disassembling of the central part of the rotor has the main advantage to not dismount the entire rotor in order to create a crack. However. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 In the third case. In regards to the 2Â rev. 7. by considering a run-down of the uncracked rotor and a rundown of a cracked one. small differences in alignment might be introduced when the central part is coupled to the other extremities.h. The measured reference situation is reported in Figs. 9. 9 and 10. the rotor presents a bow which generates around 15 mm at very low speed in bearing 1. of the element. Reference situation Even if the so-called reference situation cannot actually be considered as the true reference situation of the same rotor in this test rig. nevertheless it has been used to calculate the vibration difference. Reference case: 1Â rev. the 2Â rev. absolute vibrations. 2Â and 3Â rev. From the analysis of the 1Â rev. The distance was of 430 mm from the main inertia disk. 9–14 for the 1Â. components. reference 40 Amplitude [µm] 30 20 10 0 100 180 Phase [degrees] 90 0 -90 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Vertical Horizontal -180 100 Fig. but also a peak at about 1100 rpm that indicates a Bearing #1. component in Figs. Moreover. In fact. 7). 4. that corresponds to element 23 (Fig. . Pennacchi et al. Moreover.h. the crack has been started from a notch and made grown up to an average depth of 33 mm. 11 and 12. critical speed is rather evident at about 1 2 of the 1st critical speed.ARTICLE IN PRESS 2126 P. 1x rev.s. This depth corresponds to 47% of the shaft of 70 mm of diameter. this leads to some difﬁculties to have a valid reference case. vibration components for bearing 1. the 1st critical speed at about 1150 rpm is recognisable. Note that the solid line represents the moments on the l.2.

reference Vertical Horizontal 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Fig. 11. component in the second bearing (about 14 mm) at low speed. vibration components for bearing 1. The relatively high value of the 2Â rev. absolute vibrations. non-linear effect of the oil ﬁlm can be recognised. Pennacchi et al. 1x rev. which remains the main component over all the speed range as is also shown by the phase trend. vibration components for bearing 2. The phase difference of 180 between horizontal and .ARTICLE IN PRESS P. absolute vibrations. reference 40 Amplitude [µm] 30 20 10 0 100 180 Phase [degrees] 90 0 -90 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Vertical Horizontal 2127 -180 100 Fig. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2. 4 Amplitude [µm] 3 2 1 0 100 180 Phase [degrees] 90 0 -90 -180 100 Bearing #1. indicates a geometrical error of the shaft (journal ovalisation) in the bearing. 2x rev. Reference case: 1Â rev. 10. Reference case: 2Â rev.

vertical components is typical for ovalisation errors. The 3Â rev. 3x rev. respectively) and is mainly due to some noise. reference 18 Amplitude [µm] 16 14 12 10 8 100 180 Phase [degrees] 90 0 -90 Vertical Horizontal 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 -180 100 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Fig. . component has very reduced amplitude in both bearings (about 1 and 4 mm. absolute vibrations. 12. absolute vibrations. 13. Bearing #1.5 1 0. vibration components for bearing 2. 2x rev.5 0 100 180 Phase [degrees] 90 0 -90 180 100 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Vertical Horizontal Fig. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2. Reference case: 2Â rev. however. Reference case: 3Â rev. a smaller peak at about 1 3 of the 1st critical speed is recognisable.ARTICLE IN PRESS 2128 P. Pennacchi et al. reference 2 Amplitude [µm] 1. vibration components for bearing 1.

Reference case: 3Â rev. reference 5 Amplitude [µm] 4. 14. Hor. 1X measured. 1X additional. Hor. absolute vibrations. Pennacchi et al. Bearing #1. vibration components for bearing 2.5 4 3. Amplitude [µm] 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] 450 800 1150 Rotating speed [rpm] 1500 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. 15.5 3 100 180 Phase [degrees] 90 0 -90 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 300 500 700 900 1100 Rotating speed [rpm] 1300 1500 Vertical Horizontal 2129 -180 100 Fig. Bearing #1. 34% slot (open crack): 1Â rev. 34% slot 60 40 20 0 100 180 Phase [degrees] 90 0 -90 -180 100 Phase [degrees] Vert. 3x rev. vibration components for bearing 1. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2.ARTICLE IN PRESS P. . Case 1. 34% slot 60 40 20 0 100 180 90 0 -90 Vert.

17. vibration components for bearing 1. Hor. 34% slot 60 40 20 0 100 180 90 0 -90 Vert. Bearing #1. Bearing #1. 34% slot 40 30 20 10 0 100 180 Phase [degrees] 90 0 -90 -180 100 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 Vert. 34% slot 40 30 20 10 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. Case 1. Hor. 2X measured. Case 1. Vert. 16. 34% slot (open crack): 2Â rev.ARTICLE IN PRESS 2130 P. 1X additional. Hor. Amplitude [µm] 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. Pennacchi et al. 2X additional. 34% slot (open crack): 1Â rev. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2. . Hor. 450 800 1150 Rotating speed [rpm] 1500 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. vibration components for bearing 2. 34% slot 60 Amplitude [µm] Amplitude [µm] 40 20 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 Bearing #2. 1X measured.

Hor. 2X additional. . 2131 Amplitude [µm] Amplitude [µm] 1500 60 40 20 0 100 180 60 40 20 0 100 180 450 800 1150 Rotating speed [rpm] 450 800 1150 Rotating speed [rpm] 1500 Phase [degrees] 90 0 -90 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 90 0 -90 -180 100 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. 18. Hor. 34% slot (open crack): 2Â rev. Hor. Bearing #2.5 Amplitude [µm] 2 1.ARTICLE IN PRESS P. Case 1. 2X measured. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2.5 Amplitude [µm] 2 1. 3X additional. vibration components for bearing 2. Hor. Pennacchi et al.5 1 0. Vert. 19.5 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert.5 1 0. 34% slot 2. 34% slot Vert. Case 1. 34% slot (open crack): 3Â rev. Bearing #1. 34% slot 2. -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. 3X measured. Bearing #1. vibration components for bearing 1.5 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. 34% slot .

in which the comparison between the 2Â rev. The M 2slot identiﬁed moment is equal to 218 N m. Therefore. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #2. while the statical bending moment M is 727 N m (see Fig.e. with an error equal to 2. 5 is used. In regards to the 1Â rev. 34% slot 8 6 4 2 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. Pennacchi et al. The depth is almost correctly estimated. Taking into account that shaft diameter D is 70 mm and element 19 length l is 65 mm. Case 1: Identiﬁcation of a 34% slot The measured and the additional vibrations in the bearings. additional vibrations will be considered in the identiﬁcation procedure. Bearing #2. i. measured additional vibration in bearing 1 are . of the rotor. the r.278.ARTICLE IN PRESS 2132 P.m. the differences between the measured vibrations in the sensor positions in the reference case and in the case of the slot.3. that corresponds in Fig. for the ﬁrst three harmonic components in the case of a 34% slot are reported in Figs. components are rather small and the peak at 1 3 of the 1st critical speed is not present in all the measured vibrations. 5 to a depth of about 32%. vibration components for bearing 2. 7) in element 19. The relative residual curve has its minimum in a position in the f. A further conﬁrmation is reported in Fig. 4. The identiﬁcation results (using the additional vibration at all the available rotating speeds) are shown in Fig. it does not present a breathing behaviour and the only signiﬁcant contribution. 20. The 3Â rev. 22. 21. Hor. Since the slot is practically an always open crack.e.h. Case 1. Hor. 3X additional. components. which is not due to the unassembly/reassembly of the rotor. 34% slot 8 6 4 2 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. 15–20. 34% slot (open crack): 3Â rev. (41) is 0.1% with respect to the overall length of the rotor. Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. 3X measured. only the 2Â rev. which is very close to the actual crack position. a signiﬁcant offset is present in both the bearing at very low rotating speeds. that indicates again a permanent bow in the rotor. is due to the rotor stiffness unsymmetry which occurs twice per revolution. of Eq.s. taking into account that in this case the M 02slot =M curve of Fig.

75 Residual: 0. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Case 1 crack 2133 1 0. Similar results are obtained for the other direction and the other bearing.9 (+180) Crack depth: 32 % Fig. Experim. 34% slot (open crack): relative residual of the fault identiﬁed from 2Â rev. component. calculated 60 40 [µm] 20 0 Calc.9 0. 34% slot (open crack): comparison of additional and calculated 2Â rev displacements in bearing 1. comp.95 0.841 Element: 19 crack residual (2 x rev.85 0. Case 2: Identiﬁcation of a 14% crack The additional vibrations in the bearings for the ﬁrst three harmonic components in this case are reported in Figs. Bearing #2. Pennacchi et al. compared with the calculated ones using the identiﬁed equivalent bending moments.) Modulus: 2. 21.8 0. Case 1.18e+002 Phase: -47. Case 1. 600 700 800 900 1000 1100 1200 1300 1400 [rpm] 180 [degrees] 90 0 -90 -180 600 700 800 900 1000 1100 1200 1300 1400 [rpm] Fig.4. . 23–28. vertical. 2X additional. 4. experimental vs.ARTICLE IN PRESS P. 22.

. 1X measured. Bearing #2.ARTICLE IN PRESS 2134 P. Case 2. slot & crack: 1Â rev. Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] 450 800 1150 Rotating speed [rpm] 1500 Fig. case 2 40 Amplitude [µm] 30 20 10 0 100 180 Phase [degrees] 90 0 -90 -180 100 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 Vert. vibration components for bearing 1. slot & crack: 1Â rev. case 2 40 30 20 10 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. Hor. Hor. case 2 40 30 20 10 0 100 180 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 Vert. Hor. 24. vibration components for bearing 2. 450 800 1150 Rotating speed [rpm] 1500 450 800 1150 Rotating speed [rpm] 1500 Fig. Bearing #1. 23. 1X additional. 1X additional. Pennacchi et al. Bearing #2. 1X measured. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. Hor. Case 2. case 2 40 Amplitude [µm] 30 20 10 0 100 180 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 Vert.

2135 -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. Bearing #1. 26. 2X additional. -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. Bearing #2. Hor. Case 2. slot & crack: 2Â rev. 2X measured. slot & crack: 2Â rev. 2X measured. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. Pennacchi et al. Hor. Hor. case 2 5 Amplitude [µm] 4 3 2 1 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] Bearing #2. vibration components for bearing 2.ARTICLE IN PRESS P. case 2 20 Amplitude [µm] 15 10 5 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 Vert. case 2 5 Amplitude [µm] 4 3 2 1 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. 25. Hor. . Case 2. vibration components for bearing 1. case 2 20 15 10 5 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. 2X additional.

case 2 2 Amplitude [µm] 1. Hor.5 0 100 180 Phase [degrees] 90 0 -90 -180 100 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 Vert. Bearing #2.5 1 0. Case 2. 3X measured. 450 800 1150 Rotating speed [rpm] 1500 450 800 1150 Rotating speed [rpm] 1500 Fig. 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. slot & crack: 3Â rev. Hor. case 2 8 Amplitude [µm] 6 4 2 0 100 180 90 0 -90 Vert.5 1 0. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. Hor. Bearing #2. Case 2. slot & crack: 3Â rev. 3X additional. . Bearing #1. vibration components for bearing 2. 28. vibration components for bearing 1. case 2 2 Amplitude [µm] 1. case 2 8 Amplitude [µm] 6 4 2 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. Pennacchi et al.ARTICLE IN PRESS 2136 P.5 0 100 180 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 Vert. 27. 3X additional. 3X measured. Hor.

experimental vs.6 % Fig. in this case the crack is not properly a real crack. component. but the two curves M 02slot =M and M 02 =M of Fig. slot & crack: 2Â rev. calculated Calc. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 crack 2137 1 0.ARTICLE IN PRESS P. 5 4 [µm] 3 2 1 0 Bearing #2. As already stressed. The results of the identiﬁcation for the 2Â rev. while the statical bending moment . Pennacchi et al.646 Element: 20 Modulus: 1. instead. Experim. due to a permanent bow of the rotor.7 0. The M 2 is equal to 19 N m. component.90e+001 Phase: -5. behaviour shows a high run-out at low speed. comp. component are reported in Fig. vertical. is rather high and constant in amplitude and phase which could indicate geometrical surface errors. 30. Case 2.) Residual: 0. nor a slot. 5 are very close to each other when the depth is small. so the curve relative to an actual crack is used.4 (+180) Crack depth: 15. Case 2. 29. The 1Â rev. 400 600 800 [rpm] 1000 1200 1400 180 [degrees] 90 0 -90 -180 400 600 800 [rpm] 1000 1200 1400 Fig. 8). 2X additional. The 3Â rev. measured and calculated vertical displacements close to bearing 2.8 0. slot & crack: relative residual of the fault identiﬁed from 2Â rev. since it is breathing only for a smaller part of its surface (see Fig.6 crack residual (2 x rev. 29.9 0.

1X additional. vibration components for bearing 1. 30 20 10 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Amplitude [µm] -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. Bearing #2. Hor. case 3 50 Vert. 40 Hor.ARTICLE IN PRESS 2138 P. Pennacchi et al. crack: 1Â rev. Case 3. Bearing #2. 1X measured. 40 Hor. 1X measured. crack: 1Â rev. case 3 50 Amplitude [µm] 40 30 20 10 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. case 3 50 Vert. 31. Hor. vibration components for bearing 2. . 30 20 10 0 100 180 Phase [degrees] 90 0 -90 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 Bearing #1. -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. 32. 1X additional. Case 3. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. case 3 50 Amplitude [µm] 40 30 20 10 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert.

Pennacchi et al. 2139 -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. crack: 2Â rev. Bearing #2. case 3 60 Amplitude [µm] Vert. Hor. Bearing #2. 33. Bearing #1. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. 2X additional. Hor. case 3 40 Amplitude [µm] 30 20 10 0 100 180 Phase [degrees] Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. crack: 2Â rev. Hor. 34. 2X measured. vibration components for bearing 2. Case 3. Case 3. case 3 40 Amplitude [µm] 30 20 10 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. . vibration components for bearing 1. Hor. case 3 60 Amplitude [µm] Vert. 2X measured. 40 40 20 20 0 100 180 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 0 100 180 Phase [degrees] 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 450 800 1150 Rotating speed [rpm] 1500 Fig. 2X additional.ARTICLE IN PRESS P.

/ Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Bearing #1. Hor. Bearing #2. 10 Bearing #2. -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. case 3 Vert. Case 3. Hor. case 3 10 Amplitude [µm] 8 6 4 2 0 100 180 Phase [degrees] 90 0 -90 Phase [degrees] 450 800 1150 Rotating speed [rpm] 1500 Vert. case 3 10 Amplitude [µm] 8 6 4 2 0 100 180 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 Vert. case 3 Vert. Pennacchi et al. 3X measured.ARTICLE IN PRESS 2140 P. 10 Amplitude [µm] Amplitude [µm] 1500 5 5 0 100 180 Phase [degrees] 450 800 1150 Rotating speed [rpm] 0 100 180 Phase [degrees] 90 0 -90 450 800 1150 Rotating speed [rpm] 1500 90 0 -90 -180 100 450 800 1150 Rotating speed [rpm] 1500 -180 100 450 800 1150 Rotating speed [rpm] 1500 Fig. crack: 3Â rev. 3X additional. vibration components for bearing 2. Case 3. 3X additional. Bearing #1. crack: 3Â rev. 3X measured. Hor. vibration components for bearing 1. . 36. 35. Hor.

44e+003 Phase: -5. 6 x 10-5 Bearing #1.4 (+180) Modulus: 3. Case 3.6 0. 38. Fig.208 Element: 24 Residual: 0.01e+002 Phase: 112.49e+001 Phase: -23. experimental vs.1 (+180) Crack depth: 47.0 % Modulus: 5.3 (+180) 2x rev. Pennacchi et al. 47% cracked shaft. 3x rev.) 1 0. calculated Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 4 [m] 2 0 200 400 600 800 [rpm] 1000 1200 1400 100 [degrees] 0 -100 200 400 600 800 [rpm] 1000 1200 1400 Fig. 37.ARTICLE IN PRESS P. Residual: 0.567 Element: 23 Residual: 0. Relative residuals of the crack identiﬁcation. 3x comp.2 1x rev.976 Element: 23 Modulus: 1. . vibration components for bearing 1. crack: comparison between simulated and experimental 1Â rev.4 0. 2x. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Crack 2141 Crack residual (1x.8 0. 1X additional.

x 10-5 Bearing #1. experimental vs. crack: comparison between simulated and experimental 1Â rev. vibration components for bearing 1. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 x 10-5 Bearing #2. 39. vibration components for bearing 2. 2X additional. 1X additional. Pennacchi et al. . calculated Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 4 [m] 2 0 200 400 600 800 [rpm] 1000 1200 1400 100 [degrees] 0 -100 200 400 600 800 [rpm] 1000 1200 1400 Fig. calculated 8 6 [m] 4 2 0 200 400 600 800 [rpm] 1000 1200 1400 100 [degrees] 0 -100 Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 200 400 600 800 [rpm] 1000 1200 1400 Fig. 40. experimental vs. Case 3. Case 3.ARTICLE IN PRESS 2142 P. crack: comparison between simulated and experimental 2Â rev.

due to the unknown bow for the 1Â rev.ARTICLE IN PRESS P. vibration components for bearing 2. of Eq. 5 to a depth of 15. Case 3. in vertical direction. component. 30 for the 2Â rev. (41) is 0. experimental vs. The identiﬁcation is not shown for 1Â and 3Â rev. 4. calculated Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 6 4 [m] 2 0 200 400 600 800 [rpm] 1000 1200 1400 100 [degrees] 0 -100 200 400 600 800 [rpm] 1000 1200 1400 Fig. component (Fig. 29).s. close to bearing 2. 31–36. component. the r. 41. Since the shaft diameter D is 70 mm and element 20 length l is 65 mm. 30 shows that the ﬁtting of the calculated results to the experimental ones is rather good. From the 1Â rev. Pennacchi et al. 31 and 32) it can be inferred that the rotor presents a permanent bow which is increased with respect to the reference rotor.5. components. the high amplitude (about 40–50 mm) of the peak at 1 2 of the 1st critical speed is clearly x 10-5 Bearing #2. the identiﬁcation can be considered successful. Fig.6%. additional vibration components. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2143 M is equal to 684 N m in element 20 (Fig. . In regards to the 2Â rev. and also its relative depth is identiﬁed with good accuracy by means of the 2Â rev.h. Case 3: Identiﬁcation of a 47% crack The additional vibrations obtained with a crack of about 47% depth of the diameter in the central section are shown in Figs. 7). crack: comparison between simulated and experimental 2Â rev. The position of the crack. and to the noise for the 3Â rev. Despite the fact that the small crack (about 14%) generates only small 2Â rev. components. 2X additional. the comparison between the experimental and the calculated values (with the identiﬁed couples) in the measuring positions are shown in Fig. that corresponds in Fig. component (Figs.026. less than 2 mm in bearing 1 and less than 5 mm in bearing 2. Finally.

This leads to a very good agreement between the experimental and the simulated behaviour for the 1Â rev. component. This result has been obtained by processing the experimental data in the following way: ﬁrst the unbalances on the disks were identiﬁed. The statical bending moment M is equal to 548 N m in element 23 (Fig. component as shown in Figs. component identiﬁes the position with a particularly reduced value of the relative residual. The crack depth is identiﬁed using only the value of the 2Â rev.8 [m] 0. component as explained later. the 1Â rev. and then the dynamical behaviour due to the unbalances only has been subtracted from measured data in order to obtain the bowinduced vibrations. In this case the identiﬁcation of the crack has been attempted not only by using the 2Â rev.2 0 200 400 600 800 1000 1200 1400 [rpm] 100 [degrees] 0 -100 200 400 600 800 1000 1200 1400 [rpm] Fig.ARTICLE IN PRESS 2144 P. components. 38 and 39.425 that corresponds in Fig. The shaft diameter D is 70 mm and element 23 length l is 53. x 10-5 Bearing #1.5 mm. 3X additional. Due to this procedure. In fact the last one has a rather high value. equivalent bending moment M 2 . component at 1 3 of the 1st critical speed. (41) is 0. so that r. vibration components for bearing 1. in which bow and unbalances have been superimposed. it was considered not pertinent to use M 1 to identify the crack depth. The 1Â rev. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 due to the crack.s. which produces also a high resonance amplitude (about 10 mm) of the 3Â rev. The results in Fig. but also by the 1Â rev. Case 3.4 0.h. 42. Pennacchi et al. 7). 5 to a depth of 47%. component identiﬁes the crack in a beam element close to the actual one (element 24 instead of 23) with an error of 1. . components. while some processing were used in order to depurate the effect of the unbalance in the 1Â rev. crack: comparison between simulated and experimental 3Â rev. experimental vs. calculated Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 1 0. which is equal to 301 N m. 37 show that the location of the crack is precisely identiﬁed by 2Â and 3Â rev.6 0. and 3Â rev. of Eq. Moreover.7% with respect to the overall length of the rotor.

However. crack: comparison between simulated and experimental 3Â rev.5 0 200 400 600 800 [rpm] 1000 1200 1400 100 [degrees] 0 -100 200 400 600 800 [rpm] 1000 1200 1400 Fig. 37 due to the scale. Case 3. Pennacchi et al. component. As concerning the 3Â rev. 5. 42 and 43 for this component. 43. Three cases of crack identiﬁcations have been presented and the results. the relative residual can be considered as good. 3X additional. the relative residual is quite high. This notwithstanding.h. The comparison is reported in Figs.ARTICLE IN PRESS P. experimental vs.9 N m that made r.077 that corresponds in Fig. Since only measurements in the bearings. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 x 10-5 Bearing #2. . 5 to a depth of 30%. calculated Analytical Vertical Analytical Horizontal Experimental (difference) Vertical Experimental (difference) Horizontal 2145 1. In regards to the 2Â rev. (41) to be equal to 0. but the most remarkable result is the exact identiﬁcation of the depth of the crack. it resulted equal to 54. conﬁrm the validity of the proposed approach. Also the simulated behaviour in Figs. but this can be explained by considering that this component is normally masked by noise. or close to the bearings. if M 3 were deemed useful to identify the crack depth. Conclusions A model-based transverse crack identiﬁcation procedure in the frequency domain has been presented and applied to experimental results obtained on a medium size test rig. of Eq. because these measuring planes are standard in industrial machines. 40 and 41 is good. component.s. vibration components for bearing 2. the method is suitable for ﬁeld applications.5 1 [m] 0. the residual curve presents a well-deﬁned minimum in correspondence of the crack even this is not so evident in Fig. are used. which are rather accurate.

Pennacchi. [7] I. [4] J. A. Journal of Sound and Vibration 254 (2) (2002) 327–366. Tokyo. On the dynamics of cracked rotors: a literature survey. J. [8] H. Sekhar. Development of an on-line rotor crack detection and monitoring system. pp. [13] S. A.C.ARTICLE IN PRESS 2146 P. Bachschmid. Z. available online on ScienceDirect. 1976. 1st Mechanical Engineering Conference. J. Wauer..A. Me References [1] J. J. Bachschmid. S.R. 1057–1064. Zou. Cumberland steam plant-cracked IP rotor coupling. A comparison of different methods for transverse-crack modelling in rotor systems. Bankert. Cracking in 500 MW LP rotor shafts. Applied Mechanics Review 43 (1) (1990) 13–18. Structural Health Monitoring. Journal of Sound and Vibration 276 (1–2) (2004) 195–214. in press. Dimarogonas. [14] I. entstanden im Bereich bis zur Schleuderdrehzahl. Pennacchi. Schmerling. Zanetta. Audebert. IFToMM Fifth International Conference on Rotor Dynamics. [21] N. Italy. Crack identiﬁcation in rotating shafts by coupled response measurements. K. On the vibration behaviour of a cracked rotor. 24 November 1987. Morassi. A survey of the dynamic behaviour of a simple rotating shaft with a transverse crack. P. . Seibold. Technomic Publ. Jack. pp. in: Fu-Kuo Chang (Ed. Journal of Sound and Vibration 195 (1) (1996) 57–73. no. Hammon.R. European Journal of Mechanics A/Solids 23 (2) (2004) 325–333. E. A. September–October 1982. [20] N.oder Generatorla ¨ ufer. G. Dilena. Journal of Sound and Vibration 160 (2) (1993) 313–332. Bertilsson. 1997. L. Hawaii. Darmstadt. [9] A. [3] G.D. 1978. [22] N. IFToMM International Conference Rotordynamic Problems in Power Plants.M. Crack identiﬁcation in a rotor system: a model-based approach. Seibold.A. Markert. March 2000.J. Use of modal representation for the supporting structure in model based fault identiﬁcation of large rotating machinery: part 1—theoretical remarks. Papadopoulos. Parameter identiﬁcation of a rotor with an open crack. Mechanical Systems and Signal Processing. Bach. [10] M. Journal of Sound and Vibration 270 (4–5) (2004) 887–902.D. American Power Conference. April 1966. Bachschmid. Vibration of cracked structures: a state of the art review. Angeli.R. Azzaro. A time domain method for the localization of cracks in rotors.N. September 1986. Meccanica 35 (6) (2000) 563–582. Dong. Identiﬁcation of multiple faults in rotor systems. Lancaster/Basel. IFToMM-JSME International Conference on Rotordynamics. Gounaris. Chicago. An on-line crack detection method for turbogenerators. ISSN 0569-0692. A. pp. Stys. Stress and Reliability in Design 111 (3) (1989) 241–250. C. [12] G. Acoustique et The authors wish to thank Division Recherche et De ´ canique Vibratoire of EDF (Electricite ´ de France) for having provided test rig data. Pennacchi et al. Maschinenschaden 50 (1977) 195–204. [16] R. GroXscha ¨ den durch Turbinen.H. 338–349.S. Agnew. Scheibel. [19] N. Bucher. Diana.S. R. Tanzi.E.). P. ISROMAC-8 Conference. Gregori. Gasch. [11] G. Vania. A. F. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 Acknowledgements ´ veloppement. Investigation of the Tennessee Valley Authority Gallation Unit No. American Power Conference. Honolulu. Chicago. J. Bachschmid. A. [18] G. Determination of the fault position in rotors for the example of a transverse crack. 385–390. 325–335. [2] A. Engineering Fracture Mechanics 69 (3) (2002) 339–352. J. Engineering Fracture Mechanics 55 (5) (1996) 831–857. S. 515–524. Acoustics. Identiﬁcation of transverse crack position and depth in rotor systems. Imam. ASME Journal of Vibration. Chen. pp. September 1998. Vania. [6] ALLIANZ Berichte. Bachschmid. Pennacchi. Winert. K. S. pp. 2 turbine rotor burst. Rome. J. De ´ p. Nilsson. Haas.M. Vania.K. A two-stage approach for enhanced diagnosis of rotating machines. Vania. Patterson. The use of antiresonances for crack detection in beams. Erhardt. [5] L. [17] A. Greco. N. P. R. [15] J.

Vania. Bachschmid. Audebert. [24] N. E. Transverse crack modeling and validation in rotor systems. Pennacchi et al. S. Mechanical Systems and Signal Processing 18 (2) (2004) 329–352 ISSN 0888-3270. International Journal of Rotating Machinery 9 (2) (2003) 113–126. Pennacchi. / Mechanical Systems and Signal Processing 20 (2006) 2112–2147 2147 [23] A. Tanzi. P. including thermal effects. Experimental and theoretical application of fault identiﬁcation measures of accuracy in rotating machine diagnostics. . P.ARTICLE IN PRESS P. Pennacchi.

- Ciclo Menstrual
- Biografía Resumida de Simón Bolívar
- LABORATORIO RESISTENCIA DE MATERIALES.pdf
- EJERCICIOS MECANICA
- Ensayo de Tracción Del Acero 1045
- Momentos Respecto de Una Recta 20 Ejercicios Resueltos en 3D
- Ensayo de Tracción Acero 1045 Expo
- Ensayo de Tracción Acero 1045 Expo
- Ensayo de Tracción Del Acero 1045
- Equilibrio Cuerpo Rigido
- 218243663 Momentos Respecto de Una Recta 20 Ejercicios Resueltos en 3D
- EJERCICIOS CINEMATICA
- Equilibrio estatica
- ejercicios estatica 3d
- Modelo Viga Bernoulli
- Laboratorio de Electricidad y Electrónica Básica
- Vector
- Elementos de La Narración
- topics for ket speaking.pdf
- Diseño de Experimentos
- Diseño de Experimentos
- Cap-11-P1-Acoples
- Revisar Hidrociclon Ingles 18-06-2014
- Operation and Maintenance Cyclone
- Conversion de Grados API a Densidad Relativa

Sign up to vote on this title

UsefulNot useful by Nini Cortes

0.0 (0)

- AFEM.Ch35
- Ch31 Shell Structures
- Shear Force Bm
- SOM NOTES (Httpelearning.vtu.Ac.inp1CV331ReloadContentPreview.htm)
- Relations Between Distributed Load Shear Force and Bending Moment
- Closed Form Seabed Interaction Models for SCR
- Vol2_Section05
- Membrane Stresses
- Dis Torsion
- shearforceandbendingmomentinbeams-100114194158-phpapp01.ppt
- ozgen_d022e7897c9cb4b728964334792c9a03
- SFD BMD
- Experimental Investigation on Internally Ring-Stiffened Joints of Offshore Platforms
- 20 m span
- UNIT 2 QB
- Crowder+-+Structural+1+Fall+2010
- eurocode-2-blogspot-com.pdf
- Chapter 6 Bending
- 06 PlateTheory 03 BendingTwisting
- AS 3600-01 RC-BM-001
- Swami Vivekananda University M.tech MD Syllabus SEM 1
- Lecture Bending of Plates
- 1963_A7 R35 S4
- Topic7x Kawabata FAST
- centroidsmomentsofinertia-121222071213-phpapp02
- Lecture6 Shear Force and Bending Moment in Beams
- 1298867_634616592982133135
- Shear Force & Bending Moment Diagram Lecture
- 1020-2020-3020.pdf
- 220F11 Bending
- A Model-based Identification Method of Transverse Cracks in Rotating Shafts Suitable for Industrial Machines