Signal analysis with numerical systems
In this context, an important line of research has been devoted to the study and development of a methodology for the correction of systematic effects and for the evaluation of the uncertainty of measurement results obtained by signal digital processing in frequency domain. The proposed solutions have been implemented on DSP processor architecture, in order to evaluate the feasibility of applying them in digital instrumentation, and to carry out an experimental metrological characterization of implemented measurement systems. Starting from the techniques of "Interpolated FFT" (IFFT) known in literature, were first developed methodologies for the evaluation of uncertainty by applying the "law of propagation of uncertainty" (ISO-GUM) through processing algorithms . The proposed solution has been applied to fault diagnosis of rotating machines using the analysis of its vibration spectrum: the spectrum of the signal acquired by an accelerometer integral with the machine was constantly compared to the spectrum that occurs in normal operation ("unfaulty "). Deviations may help to early diagnose the onset of failures, and the analysis of the signal in the frequency domain allows to classify the fault into one of the possible fault classes provided. The developed system was also characterized in terms of response time. Subsequent phases of the research activities were directed to the development of techniques for the correction of the effects of harmonic interference on the estimation of parameters of a spectral component due to the other spectral components of the signal. A procedure called "IFFTc" for the correction of the effects of harmonic interference was developed, implemented on a DSP architecture and characterized. The procedure for the assessment of the uncertainty of parameter estimations has been developed also for the IFFTc algorithm. Finally, the research has been devoted to the development of a procedure for the relief in the spectral amplitude of any "hidden tones", that is, spectral components which are not detected as local maxima of the amplitude spectrum since they are hidden by close components with greater amplitude. A further research in this area concerns the development of a method for the detection of very low frequency spectral components in the signals due to the suspension displacement on motorbikes equipped with magneto-rheological suspension systems.
Involved researchers:
