**4. Condition monitoring of I.C. engines in cold conditions**

This second application addresses the use of CWT and DWTs as a means of quality control for assembly faults in diesel engines using cold test technology. Nowadays, the majority of engine manufacturers test their engines by means of "hot tests", i.e. tests in which the engine is firing. Hot tests are mainly aimed at determining engine performance.

Recently, some companies have introduced "cold tests" which aim to identify assembly anomalies by means of torque, pressure and vibration measurements. Cold tests are more oriented towards identifying the source of anomalies since they are not affected by noise and vibration due to firing. Reciprocating machines, such as IC engines, give non-stationary vibration signals due to changes in pressure and inertial forces and valve operations. Therefore, WTs are an efficient tool for analyzing transient events during the entire engine operation cycle.

Here, CWTs are applied in order to obtain an accurate fault event identification for signals measured from engines with different assembly faults that have not been considered in literature. The analysis takes advantage of cyclostationary modelling developed and tested by Antoni in [8].

Experimental investigations were carried out on a 2.8 dm3 4-cylinder 4-stroke, four-valve-percylinder turbocharged diesel engine with an exhaust-driven turbo-compressor produced by VM Motori. The measurements were carried out in cold conditions (without combustion) while the engine crankshaft was driven by an electric motor via a coupling. The acceleration signal was measured by means of a piezoelectric general purpose accelerometer mounted on the engine block (turbocharger side) close to the bearing support of the crankshaft. A 360 pulse/rev tachometer signal was used to measure the angular position of the crankshaft. During acquisition, the acceleration signal was resampled with a 1 degree angular resolution.

a b

Fig. 7. Continuous wavelet transform (Mexican hat wavelet) for transom left-side acceleration, 700-1500 Hz frequency range; Coupling: Type 1 (a) and Type 2 (b).

**4. Condition monitoring of I.C. engines in cold conditions** 

impulse signal.

by Antoni in [8].

Moreover, a comparison between two different Morlet and Mexican hat wavelet functions was evaluated. Undeniably, the Mexican hat wavelet function has a shape which is totally inadequate for analysing the signal impulse components. This is shown by the results in Fig. 7 which indicate that the Mexican hat highlighted the different frequency contempt of two coupling types but was not able to precisely localise the higher frequency components of the

This second application addresses the use of CWT and DWTs as a means of quality control for assembly faults in diesel engines using cold test technology. Nowadays, the majority of engine manufacturers test their engines by means of "hot tests", i.e. tests in which the

Recently, some companies have introduced "cold tests" which aim to identify assembly anomalies by means of torque, pressure and vibration measurements. Cold tests are more oriented towards identifying the source of anomalies since they are not affected by noise and vibration due to firing. Reciprocating machines, such as IC engines, give non-stationary vibration signals due to changes in pressure and inertial forces and valve operations. Therefore, WTs are

Here, CWTs are applied in order to obtain an accurate fault event identification for signals measured from engines with different assembly faults that have not been considered in literature. The analysis takes advantage of cyclostationary modelling developed and tested

Experimental investigations were carried out on a 2.8 dm3 4-cylinder 4-stroke, four-valve-percylinder turbocharged diesel engine with an exhaust-driven turbo-compressor produced by VM Motori. The measurements were carried out in cold conditions (without combustion) while the engine crankshaft was driven by an electric motor via a coupling. The acceleration signal was measured by means of a piezoelectric general purpose accelerometer mounted on the engine block (turbocharger side) close to the bearing support of the crankshaft. A 360 pulse/rev tachometer signal was used to measure the angular position of the crankshaft. During acquisition, the acceleration signal was resampled with a 1 degree angular resolution.

engine is firing. Hot tests are mainly aimed at determining engine performance.

an efficient tool for analyzing transient events during the entire engine operation cycle.

The first faulty condition concerned an engine with a connecting rod with incorrectly tightened screws, that is, screws which were only tightened with a preload of 3 kgm, instead of the correct torque of 9. The second faulty condition concerned an engine with an inverted piston, with incorrectly positioned valve sites. This incorrect assembly hindered the correct correspondence between the valve plates and the valve sites. Since the exhaust valve site area is larger than the intake valve site, the exhaust valves knocked against the noncorrespondent intake valve sites.

Fig. 8(a) shows that the CWT map (Impulse wavelet) of the Time Synchronous Average (TSA) detected four cylinder pressurizations and two events related to the faulty condition. Even if a remarkable vertical line at 100 degrees was present in the CWT map of the TSA (Fig. 8 (a)), it is not sufficient to assure the presence of a mechanical fault since its amplitude is comparable to the pressurization peak amplitudes. Therefore, the CWT of the residual signal (i.e. the signal obtained by subtracting the time synchronous average from the raw signal) is an expected step in mechanical fault localization within engine kinematics (Fig. 8(b)). As depicted in Fig. 8(b), the presence of the pre-loaded rod is highlighted by a marked vertical line at about 100º.

As explained in [32] the peak is caused by the absence of controlled bush deformation when the correct tightening torque is not applied. This clearance is abruptly traversed whenever a change in the direction of the resultant force occurs on the rod. In particular, it was demonstrated that the acceleration peak took place at the beginning of the cylinder 3 intake stroke, corresponding to cylinder 2 pressurization (i.e. 'Press 2' in Fig. 8(a)). Hence, fault location can be only achieved by the analysis of the residual signal. It is worth noting that better angular fault localization can be achieved using the Morlet mother wavelet (Fig. 8(c)) which gives lower frequency resolution but higher angular localization of the anglefrequency map. Since the purpose of the proposed approach is to obtain reliable fault diagnostics through accurate angular transient event localization, the Morlet wavelet can be considered the most desirable if compared with the Impulse wavelet.

In order to improve the CWT of the TSA, the purification method was firstly carried out using correlation weighted CWT coefficients, i.e. ˆ *CWT* , as described in Section 2.1.1.

As previously mentioned, the correlation coefficient (,) *a t* used in this method is able to select which coefficient gives the best match between the frequency of the signal and the frequency corresponding to the Impulse wavelet scale.

Fig. 9(a) shows that this method provides a clearer representation in terms of sensitivity to background noise. However, the use of the coefficient correlation method does not improve the angular localization of the main engine events. As noted earlier, this enhancement can be obtained using the Morlet mother wavelet. The Morlet mother wavelet was used to compute the wavelet transform by means of both traditional and TDAS methods. No significant improvements in angular faulty localization can be obtained by using the TDAS method (Fig. 9(b)). Therefore, it can be concluded that a traditional CWT map with a Morlet mother wavelet is sufficient for faulty localization purposes.

It should be noted that CWT is used in order to distinguish faulty conditions from normal ones for practical fault diagnosis and not to obtain reliable parameters for an automatic procedure led by a data acquisition system.

In order to overcome this issue, the DWT technique for the extraction of faulty components from the signal, proposed by Shibata, was evaluated for the second fault which was condition tested, i.e. the inverted piston.

On the Use of Wavelet Transform for Practical Condition Monitoring Issues 367

Fig. 9. Faulty engine – (a) CWT of the TSA: purification method (impulse mother wavelet);

b

a b

a

Fig. 10. DWT coefficients for the vibration signals (120 rpm): (a) Normal condition; faulty

(b) TDAS method (morlet mother wavelet).

condition (piston inverted).

Fig. 10 shows the DWT coefficients ( *j k*, *c* ) when Symlet (eight order) is used for the wavelet and the scaling function. Data sampled at 70 μs were used for the DWT.

Fig. 8. Faulty engine – (a) CWT (impulse wavelet) of the TSA, (b) residual signal (impulse wavelet); (c) CWT (morlet wavelet) of the residual signal.

Fig. 10 shows the DWT coefficients ( *j k*, *c* ) when Symlet (eight order) is used for the wavelet

b

c

a

Fig. 8. Faulty engine – (a) CWT (impulse wavelet) of the TSA, (b) residual signal (impulse

wavelet); (c) CWT (morlet wavelet) of the residual signal.

and the scaling function. Data sampled at 70 μs were used for the DWT.

Fig. 9. Faulty engine – (a) CWT of the TSA: purification method (impulse mother wavelet); (b) TDAS method (morlet mother wavelet).

Fig. 10. DWT coefficients for the vibration signals (120 rpm): (a) Normal condition; faulty condition (piston inverted).

On the Use of Wavelet Transform for Practical Condition Monitoring Issues 369

[1] Papoulis, A., 1962, The Fourier Integral and its applications. McGraw-Hill, New York. [2] Van der Auweraer, H., et al., (1992), Spectral estimation of time-variant signal, in

[3] Van der Auweraer, H., et al., (1992), Analysis of non-stationary noise and vibration

[4] Bartelmus, W., (2001), Mathematical modelling and computer simulations as an aid to gearbox diagnostics, *Mechanical Systems and Signal Processing*, 15, 855-871. [5] Jid, S., Howard, I., (2006) Comparison of localized spalling and crack damage from

[6] Wu, J.D. and Chuang, C.Q. ,(2005). "Fault diagnosis of internal combustion engines

[7] Shibata, K., Takahashi, A., and Shirai, T. (2000). Fault diagnosis of rotating machinery

[8] Antoni, J., Daniere, J., and Guillet, G. (2002), Effective vibration analysis of ic engines

[9] Da Wu, J., Chen Chen, J. (2006), Continuous wavelet transform technique for fault signal diagnosis of internal combustion engines, *NDT&E International*, 39, 304-311. [10] Tse, P., Yang W., Tam, H. Y., (2004), Machine fault diagnosis through an effective exact wavelet analysis,, *Mechanical Systems and Signal Processing*, 277, 1005-10024. [11] Geng, Chen, J., Barry Hull, J. ,(2003). Analysis of engine vibration and design of an

[12] Farag K. Omar, A.M. Gaouda (2012), Dynamic wavelet-based tool for gearbox

[13] Loutas, T. H., Roulias, D., Pauly, E., Kostopoulos, V. (2010). The combined use of

[14] Torrence, C., (1998), A Pratical Guide to Wavelet Analysis. Bulletin of the American

[15] Peng, Z.,K., Chu, F. L., (2004), Application of the wavelet transform in machine

[16] Al-Badour, F., Sunar, M., Cheded, L. (2011), Vibration analysis of rotating machinery

[17] Newland, E., 1994, Wavelet Analysis, Part I: Theory, *Journal of Sound and Vibration* 116,

diagnosis, *Mechanical Systems and Signal Processing*, 26, 190-204.

*Proceedings of ISMA17 International Conference on Noise and Vibration Engineering*,

signals, in Proceedings of ISMA17 International Conference on Noise and Vibration

dynamic modelling of spur gears vibrations, *Mechanical Systems and Signal* 

using visual dot patterns of acoustic and vibration signals", *NDT&E International*,

through visualisation of sound signal, *Mechanical Systems and Signal Processing*, 14,

using cyclostationarity. Part I-A methodology for condition monitoring, *Journal of* 

applicable diagnosing approach, *International Journal of Mechanical Sciences*, 45,

vibration, acoustic emission and oil debris on-line monitoring towards a more effective condition monitoring of rotating machinery, *Mechanical Systems and Signal* 

condition monitoring and fault diagnostics: a review with bibliography, *Mechanical* 

using time–frequency analysis and wavelet techniques, *Mechanical Systems and* 

**7. References** 

Leuven, Belgium, pp. 207-223.

*Processing*, 20332-349.

38(2005), pp. 605-614.

*Sound and Vibration*, 257, 815-837.

229-241.

1391-1410.

409-416.

*Processing*, 25, 1339-1352.

Meteorological Society, 79(1).

*Signal Processing*, 25, 2083-2101.

*Systems and Signal Processing* 18, 199-221.

Engineering, Leuven, Belgium, pp. 385-405.


Table 1. Mean value of the peaks of acceleration (m/s2).


Table 2. Comparison with coefficients of DWT: ratio of RMS value between the faulty (F) and normal (N) conditions.

Table 2 shows the comparison between the RMS ratio of the DWT coefficients in faulty (F) and normal (N) conditions with the engine running at 120 rpm. The j = 1 level shows the highest difference between the faulty and normal vibration signals. Thus, the RMS ratio at the first decomposition level may be considered a reliable monitoring feature.
