Statistical parameters of vibration data distributions provide a remarkable source of information in condition monitoring. The assumption of normality leads to comparatively simple description within the framework of classical statistics. However, due to the presence of outliers and heavy-tailed distributions, this approach is often unacceptable. In such cases, robust methods prove superior. The paper is focused on applying robust statistics in analyzing vibration data dispersion and correlation for the purpose of lifetime consumption estimation. This approach is suitable for large rotating machines operated in an industrial plant environment and characterized by significant influences of control parameters and interference. It is shown that robust meth¬ods yield results that are easier to interpret from both qualitative and quantitative viewpoints. Examples re¬ferring to large steam turbines operated by utility power plants are given; however, certain results can be generalized over a broader class of rotating machines or even diagnostic objects.
Vibration-based symptoms are extremely important in monitoring technical
condition of virtually all types of rotating machinery, especially large and
complex units, e.g. turbo-generators, large fans or compressors. This is
justified by their high information content, non-intrusive nature and
comparatively well-developed data acquisition and processing procedures. The
most straightforward approach consists in analyzing vibration patterns recorded
at a certain moment on the basis of a purpose-developed diagnostic model.
Information on the object condition is, however, also contained in time
histories of certain vibration components. As pointed out in References, a
vibration component evolution type and relevant timescale (the latter varying
within a broad range, from seconds to months) contain information on a fault
type and thus are useful already at the qualitative diagnosis stage. More
detailed analysis of vibration time histories becomes even more important when
it comes to quantitative assessment and is mandatory for a prognosis.
Morel J., “Vibration des Machines et Diagnostic de Leur État
Mécanique”, Eyrolles, Paris, France, 1992.
Orłowski Z., “Diagnostics in the Life of Steam Turbines”, WNT,
Warszawa, Poland, 2001 (in Polish).
Natke H. G. and Cempel C., “Model-Aided Diagnosis of Mechanical
Systems”, Springer, Ber-lin-Heidelberg-New York, 1997
Gałka T., “Influence of Load and Interference in Vibration-Based
Diagnostics of Rotating Machines”, Advances and Applications in Mechanical
Engineering and Technology, Vol 3, No 1, pp 1-19, 2011.
Gałka T. and Tabaszewski M., “An Application of Statistical
Symptoms in Machine Condition Diagnostics”, Mechanical Systems and Signal
Processing, Vol 25, No 1, pp 253-265, 2011.
Maronna R. A., Martin R. D. and Yohai V. J., “Robust Statistics”,
Wiley, Chichester, UK, 2006
Gałka T., “Diagnostics of the steam turbine fluid-flow system
condition on the basis of vibration trends analysis”, Proceedings of the
7th European Conference on Turbomachinery, Athens, Greece, pp
521-530, March 2007.
Gałka T., “Statistical Vibration-Based Symptoms in Rotating
Machinery Diagnostics”, Diagnostyka, Vol 2, No 46, pp 25-32, 2008.
Gałka T., “Application of Energy Processor Model for Diagnostic
Symptom Limit Value Determination in Steam Turbines, Mechanical Systems and
Signal Processing”, Vol 13, No 5, pp 757-764, 1999.
Gałka T., “The ‘Old Man Syndrome’ in machine lifetime consumption
assessment”, Proceedings of the CM2011/MFPT2011 Conference, Cardiff, UK,
paper No 108, June 2011.
Gałka T., “Assessment of the turbine fluid-flow system condition
on the basis of vibration-related symptoms”, Proceedings of the 16th
International Congress COMADEM 2003, Växjö University, Sweden, pp 165-173,
Upton G. and Cook I., “Understanding Statistics”, Oxford
University Press, UK, 1996.
Aldrich J., “Correlations Genuine and Spurious in Pearson and
Yule”, Statistical Science, Vol 10, No 4, pp 364-376, 1995.
Rousseeuw P. J. and Croux C., “Alternatives to Median Absolute
Deviation”, Journal of the American Statistical Association, Vol 88, pp
Martin R. D. and Zamar R. H., “Bias Robust Estimates of Scale”,
The Annals of Statistics, Vol 21 pp 991-1017, 1993.
Rodgers J. L. and Nicewander W. A., “Thirteen Ways to Look at the
Correlation Coefficient”, The American Statistician, Vol 42, No 1, pp 59-66,
Abdi H., “The Kendall Rank Correlation Coefficient”, in
“Encyclopedia of Measurement and Statistics”, N Salkind (Ed.), Sage, USA,
Scheaffer R. L. and McClave J. T., “Probability and Statistics for
Engineers”, PWS-KENT Publishers, Boston, USA, 1986.
Bachschmid N., Pennacchi P. and Tanzi E., “Cracked Rotors. A
Survey on Static and Dynamic Behaviour Including Modelling and Diagnosis”,
Springer, Berlin, Germany, 2010.
Bently D. E. and Hatch C. T., “Fundamentals of Rotating Machinery
Diagnostics”, Bently Pressurized Bearing Press, Minden, USA, 2002.
Gałka T., “Correlation-based symptoms in rotating machines
diagnostics”, Proceedings of the 21st International
Congress COMADEM 2008, Praha, Czech Republic, pp 213-226, 2008.
Gałka T., “Correlation Analysis in Steam Turbine Malfunction
Diagnostics”, Diagnostyka, Vol 2, No 54, pp 41-50, 2010.
Gałka T., “Application of the Singular Value Decomposition method
in steam turbine diagnostics”, Proceedings of the CM2010/MFPT2010 Conference,
Stratford-upon-Avon, UK, paper No 107, 2010.
Sethuraman J., “The Asymptotic Distribution of the Rênyi Maximal
Correlation”, Florida State University Technical Report No M-835, 1990.
Strickert M., Schleif F-M., Seiffert U. and Villmann T.,
“Derivatives of Pearson Correlation for Gradient-Based Analysis of Biomedical
Data”, Revista Iberoamericana de Inteligencia Artifical, Vol 12, No 37, pp
Székely G. J. and Rizzo M. L., “Brownian Distance Covariance”, The
Annals of Applied Statistics, Vol 3, No 4, pp 1236-1265, 2009.
Tukey J. W., “Useable resistant/robust techniques of analysis”,
Proceedings of the First ERDA Symposium, Los Alamos, USA, pp 11-31, 1975.