Spectral Analysis of a Rotating Machine Laboratory

Abstract

Spectral analysis has shown to be essential in many current fields as the data from the signals collected has been utilized for various reasons. The analysis has found use in the health sectors, seismological fields, and in control systems, among others. The report aimed at investigating the spectral analysis for rotating parts of a machine. Various software was utilized to achieve the aim and objectives being investigated. The essential tools utilized involved the spectral analysis of the SignalCalc Ace Dynamic capabilities of the analyzer and the accelerometer (PCB) that was to be powered by the platform of Quattro. These were software with the support hardware of a four-input to one output for the diagnosis of rotating machine, vibration testing, and structural vibration. The focus was, therefore, on the practical skills development of the vibration analysis experiment for sophisticated systems and the critical appreciation ability of the effect of the processing elements for a precise analysis. From the results, the indication was that different input parameters affected the amplitude and frequency of the outputs. This, in return, affected the accuracy of the analysis as small changes in the input data produced adversely different outputs for the various tests being investigated.

Contents

Abstract 2

1. Introduction. 4

1.1.     Aim.. 5

1.2.     Objectives. 5

2. Theory. 5

2.1.     Apparatus. 6

3. Procedure. 7
4. Results and Discussion. 8
5. Conclusion. 15

References. 17

Spectral Analysis of a RotatingMachine Laboratory

1.      Introduction

The applications of spectral analysis are diverse in different fields of studies. In the case of monitoring vibrations, the content of spectral for the measured signals provide information for the wear together with other mechanical parts characteristics being considered. In the field of economics, astronomy, and meteorology, the analysis of spectra could illustrate periodicities that are hidden in the data being studied that are related to the recurring processes of cyclic behavior. In the analysis of speech, the models involved in the voice signals are essential for the comprehension of the process of speech production. Also, they can be utilized in both speech recognition and speech synthesis. Spectral analysis has also been utilized in sonar and radar systems as the signal received gives information on different locations of the targets producing the signals in the view field. In the health field, the analysis has proven useful as the measured signals from the patients like the ECG and in EEG signals give essential information on the diagnosis materials. Again, spectral analysis has many uses in seismology in that the recorded signals before and at the time of seismic phenomena like earthquakes and volcanic eruption provides vital information on the movement of the ground as it relates to the phenomena assisting in prediction of the said events. The estimation of seismic spectral is also utilized in subsurface geological structure prediction in the exploration of oil and gas. Additionally, the analysis finds useful in control systems as there is an upcoming need as a means of characterization for behavior (dynamical) of an involved system and eventually synthesize the system’s control.

1.1.Aim

The experiment aimed at the spectral analysis for the dynamic analysis of a rotating machine while indicating the approaches of determining and identifying rotational components influencing the processing of the parameters needed for accuracy of frequency response.

1.2.Objectives

1. Identification of rotational speed peaks
2. Assessment of windowing effect on the involved signal
3. Assessment of the influence of spectral representation processing resolution
4. Application of zoom function for accuracy testing for the spectral representation peak

2.      Theory

Modal and spectral analysis are common methods for analysis utilized in rotating machines diagnostic, noise analysis, and structural testing. The final aim of the two is a simplification of sophisticated responses in vibration to the primary elements, which make it. The basis of the analysis is the harmonic motion that is the simplest form of vibration for a signal associated with the mass-spring system vibration, rotating components, and simple pendulum. The representation of harmonic is a sinusoidal motion of a given location. The following equation can define harmonic motion:

In that A is the amplitude, ω is the angular frequency given as 2лf in rad/sec.

For a complex harmonic, each of the fan parts produces a harmonic component with a specific amplitude and frequency. Whenever all the parts are in rotation, a time signal is a result, which is made up of each element’s harmonic characteristics. For such cases, it is impossible to provide adequate data from time signal only. However, the analysis (spectral) assists in the identification of the amplitudes and frequencies for the involved rotating elements. The following equation could be utilized for N element parts on a harmonic motion for a complex vibration.

In that Ai is the amplitudes,  is the frequencies, and  as the phases for the elements making up the signal involved. It is essential to note that the basis of spectral analysis is the Fourier theorem given by the following equation for all periodic signal;

This can also be in the form of summation cosinusoidal and sinusoidal components at similarly spaced elements of 1/nT as illustrated in the following equation;

sin 𝑛2𝜋𝑓𝑡)

In that,  and bnsin 𝑛2𝜋𝑓𝑡 represent the signal harmonics.

2.1.Apparatus

• DataPhysics Quattro as the acquisition equipment
• Accelerometer (4 PCB) this represented an electromechanical component for measuring forces due to acceleration with a rand of about 0.064 ft/sec² depending on the clock speed and configuration
• DataPhysicsSignalCalc Ace as the analysis software

3.      Procedure

1. The SignalCalc software for analysis was opened and from the option selected AutoSpectrum
2. Only two of the four channels (Quattro) were utilized as a tick was left for the first two of them.

• Again, the accelerometer (PCB piezoelectric) was utilized, changing the coupling into ICP of 2 mA for the channels in procedure II above. The calibration data for the various data was introduced since the units for acceleration were in m/s², and calibration charts were identified. The EU values were then converted from V into m/s^2, and the sensitivity values for mV/EU were introduced from the selected charts for calibrations.

1. The acquisition parameters were then adjusted. Necessary parameters were provided for the processing and acquisition of accuracy for the data. The range required was to a maximum of 500 Hz, the maximum frequency was set at 400 Hz with spectral lines of 800 lines. The averages number was changed to twenty, and the window was left as Hanning.
2. The plot scale was then adjusted accordingly then one of the graphs was righted clicked to choose attributes for the graph while deselecting dB for the log. This was repeated for the second graph involved.
3. The test was then started while identifying the amplitudes for harmonics from the plot. The fan was switched on from the right-hand side, ensuring that the amplifiers’ voltage was always less than ten volts. From the graph, the cursor was moved to illustrate the peak value and the harmonic components.

• A report was then created from the collected data and saved.
• The voltage was then reduced to get the new form of the spectrum, and the peak value noted.

1. A complex spectrum was then tested, and the new spectrum observed.
2. The spectral lines were changed utilizing high and low resolutions with the number changing from eight hundred to four hundred and saved the graphs.
3. Window changing was then carried out in that there was an assessment of windowing influence on the representation of the signal. The change was to Rect fromHanning, and the graphs were saved. The measurements were then restarted, and the new results observed.

• Zoom was selected for twenty Hertz for Fct and 20 Hz for the frequency span.

4.      Results and Discussion

The results of the experiment were as illustrated below.

Figure 1. Right fan 8 v, frequency 29.5 Hz, amplitude 1.08 mm/s².

From Figure 1 above, the frequency was observed to be 29.5 at a voltage of 8 v for the right fan. The amplitude at this voltage was observed at 1.08 mm/s²for the Hanning window. The left fan results were as illustrated in Figure 2 below, which recorded a frequency of 0 Hz with an amplitude of 11 mm/s².

Figure 2. Left fan 8 v, frequency 0 Hz, amplitude 11 mm/s².

The frequency was at zero at the origin of the plot or at the left fan under the same 8 v, as can be observed in Figure 2 above. At the lowest frequency, the amplitude was at peak while at the highest frequency, the amplitude was least as can be seen in Figures 1 and 2.

Figure 3. Right fan 5.5 v, frequency 19 Hz, amplitude 300 mm/s².

Reducing the fan voltage reduces the frequency while increases the amplitude of the plot. Voltage is directly proportional to frequency while inversely proportional to the amplitude of the plot, as can be illustrated in Figure 3 above.

Figure 4. Left fan 5.5 v, frequency Hz, amplitude 10.7 mm/s².

There was a little effect with a reduction of the voltage on the left fan, as can be observed in Figure 4 above.

Figure 5. Right fan 8 v, frequency 30 Hz, amplitude 849 mm/s².

From Figure 5, it can be seen that the frequency increased together with the amplitude on the Hanning window. The peak amplitude was valued at 849 mm/s² with a frequency of 30 Hz.

Figure 6. Left fan about 9 v, frequency 15 Hz, amplitude 254 mm/s².

Involving a voltage of about 9 v resulted in doubling the speed at the right fan compared to the left fan. With additional voltage on the right fan, the result is a high frequency, and frequency is directly proportional to speed, as explained in theory.

Figure 7. Right fan 8 v, frequency 30 Hz, amplitude 254 mm/s².

Figure 7 above illustrates the plot with a reduction of spectral lines from 800 to 200. The effect was in an increase of frequency since the frequency and the spectral lines are inversely proportional to the same voltage.

Figure 8. Left fan about 9 v, frequency 18 Hz, amplitude 617 mm/s².

With an increase in voltage at the same spectral lines, the effect was an increase of both the amplitude and the frequency, as illustrated in Figure 8 above.

On changing the window from Hanning to Rect, the following results were obtained.

Figure 9. Right fan 8 v, frequency 30 Hz, amplitude 1.23 mm/s².

A comparison of the two gave slightly different results, and this was a result of the difference in sensitivity and range of the windows. This can be seen in Figure 1 as compared to Figure 9.

Figure 10. Right fan 9 v, frequency 18 Hz, amplitude 629 mm/s².

The case was the same with an increase in voltage as compared between Figure 2 and Figure 10 above.

Utilizing zoom Fct 30 Hz, F span 50, and spectral lines of 800 Figure 11 below illustrated the results.

Figure 11. Zoom Fct 30 Hz, F span 50, and spectral lines of 800

5.      Conclusion

The aim of the experiment was meant together with the stated objectives. It was observed that changing the numerical input values had an influence on the sampling time together with the frequency resolution. The higher the amplitude, the lower was the frequency since they are inversely proportionate. Speed was also influenced by changes in voltages; therefore, changes in frequency and amplitude in that the higher they were, the greater was the speed. Various materials have different frequencies, and the data obtained could be useful in the study of various material properties, as has been illustrated. Sensitivity and range differences had an impact on the final results found in the data collected in terms of accuracy.

References

Abrahamsson, R., Jakobsson, A., & Stoica, P. (2004). Spatial Amplitude and Phase Estimation Method for Arbitrary Array Geometries. Sweden: Uppsala University.

Angelo, M. (1987). Vibration Monitoring of Machines. Bruel&Kjaer Technical Reviews.

Baysal, U., & Moses, R. (2003). On the geometry of isotropic arrays. IEEE Transactions on Signal Processing, 1469–1478.

Bloomfield, P. (1976). Fourier Analysis of Time Series- An Introduction. New York: Wiley.

Bracewell, R. N. (1986). The Fourier Transform and its Applications, 2nd Edition. New York: McGraw-Hill.

Brockwell, R. J., & Davis, R. A. (1991). Time Series- Theory and Methods, 2nd Edition. New York: Springer-Verlag.

Broersen, P. M. (2002). Automatic spectral analysis with time series models. IEEE Transactions on Instrumentation and Measurement, 211-216.

Dossing, O. (1988). Structural Testing, Part I: Mechanical Mobility Measurements. Bruel&Kjaer Publications.

Gade, S., & Herlufsen, H. (1987). Use of Weighting Functions in DFT/FFT Analysis (Part I). Bruel&Kjaer Technical Reviews.

Haykin, S. (1995). Advances in Spectrum Analysis and Array Processing, Volume 3. New Jersey: Prentice-Hall.

McKelvey, T., & Viberg, M. (2001). A robust frequency domain subspace algorithm for multi-component harmonic retrieval. Proceedings of the 35th Asilomar Conference on Signals, Systems, and Computers (pp. 68–72). California: Pacific Grove.

Ninness, B. (2003). The asymptotic CRLB for the spectrum of ARMA processes. IEEE Transactions on Signal Processing, 1520–1531.

Proakis, J. G., Rader, C. M., Ling, F., & Nikias, C. L. (1992). Advanced Digital Signal Processing. New York: Macmillan.

Rao, S. (2011). Mechanical Vibrations. Singapore: Prentice Hall.

Stoica, P., & Larsson, E. G. (2001). Comments on the ‘Linearization method for finding Cram´er-Rao bounds in signal processing. IEEE Transactions on Signal Processing, 3168–3169.

Stoica, P., Larsson, E. G., & Gershman, A. B. (2001). The stochastic CRB for array processing: a textbook derivation. IEEE Signal Processing Letters, 148-150.

Stoica, P., Larsson, E. G., & Li, J. (2000). Adaptive filter-bank approach to restoration and spectral analysis of gapped data. The Astronomical Journal, 2163–2173.

Thompson, W. T. (1998). Theory of Vibration with applications. N.J.: Prentice-Hall.

Trees, H. L. (2002). Optimum Array Processing (Part IV of Detection, Estimation, and Modulation Theory). NY: Wiley.

Wei, W. (1990). Time Series Analysis. NY: Addison-Wesley.