Envelope curve spectrum
1.1. Relationship between time signal and frequency signal
With all condition monitoring sensors from Balluff, time domain evaluations such as RMS, peak, magnitude, kurtosis... can be evaluated.
The following illustration shows a vibration signal with 1Hz vibration frequency and an amplitude of 3.
If a spectrum (frequency signal evaluation) is formed from the time domain evaluation by the FFT, the signal looks like below.
And here lies the difference between the BCM variants. Only the BCM0003 can evaluate the spectrum.
Hir is a display error. The amplitude should also be 3.
Here again summarized in a representation:
1.2. Spectrum of an oscillation
In the last chapter, the relationship between a time signal and a subsequent frequency signal was illustrated using one signal.
However, in practice it is far more complex and the time signal already has a proportion of many signals (red signal in the illustration).
The frequencies contained in the red signal are shown with the different proportions in blue, purple and green.
A correlation can be seen by looking at the "frequency domain" on the right-hand side.
The amplitude of the frequencies and their individual frequencies are easy to recognize, which is somewhat difficult in the time domain. It is also easy to see how many different frequencies are contained in a signal.
An example with two frequency signals is considered. These two frequencies can also be clearly seen in the time signal (next image).
These frequencies were used for illustration and simplification.
- Signal = 1Hz and an amplitude of 3
- Signal = 16Hz and an amplitude of 1
The amplitudes and their heights are quite easy to recognize. The individual frequencies cannot be easily extracted from the time domain display.
If you now look at the frequency signal rather than the time signal, you can see the individual frequencies with the amplitudes of the signals very clearly.
Here again summarized in a representation:
2. Why do you need the envelope spectrum?
The envelope spectrum makes additional frequencies "visible" that cannot be detected with a pure FFT. In order to be able to analyze the other frequencies, the envelope curve is placed around the time signal in the envelope curve analysis and only then is the Fourier transformation carried out.
If there is an imbalance or resonance vibrations of the entire system on the vibration to be analyzed, the envelope analysis has the advantage of "hiding" these. This is why it is used to good effect in the monitoring of roller bearings.
2.1. Recognizing bearing/ toothed pulley damage
These shock pulses and therefore their frequency are dependent on the rotational frequency of the bearing. Consequently, the frequency of the shock pulses changes when the rotational frequency of the bearing changes.
3.1. High-pass filters and mathematical operations
The time signal, which can also contain other machine noises, imbalances and interference signals, must first of all be filtered (high-pass filter or band-pass filter) and processed using mathematical operations.
The picture shows the signal before the filtering process.
A signal is then obtained that shows the "clocking of the bearing damage". The resonance frequency was not evaluated by the filter. This example shows damage in the outer ring of the rolling bearing.
3.2. Rectification of the raw signal and formation of the envelope curve
The signal obtained from the bearing must now be rectified in order to obtain an appropriate envelope curve from this signal.
The graphic shows a rectified signal with an envelope curve around it.
3.3. FFT of the envelope curve
In an FFT of a modulated oscillation, only the carrier frequency and its sidebands are displayed at the distance of the modulation frequency. This can be seen in the next graphic.
If the FFT is applied to the envelope signal, you will not see any carrier frequency in the next graphic. Only the modulation frequency of the envelope is determined directly. This is how the frequency of the shock pulses is determined in an envelope analysis.
4.1. General settings
The following settings can be made for the envelope curve analysis in the envelope curve configurations.
The data points of a spectrum are constant at 1714 data points per spectrum.
The image below shows the use of a filter. This has already been mentioned in previous chapters. As a high-pass filter makes the most sense for envelope analysis, the lower band limit can be set to a minimum of 1000Hz.
When selecting the spectrum range, it depends on the spectrum resolution that you want to achieve. It also depends on the speed (see table above).
The averaging function can be used to average several spectra together in order to suppress spikes or one-off events. In the image shown below, 8 spectra are averaged.
By averaging 8 in this example, the acquisition time is increased accordingly. By averaging 8 spectra, the acquisition time of the table shown above of 286ms at 6000Hz spectrum range is multiplied by 8.
The spectrum resolution remains the same.
A distinction is made between two settings in the band mode settings.
More on this in the next chapters.
4.2. Multipliers for the rotational speed
Multipliers for the rotational speed are to be used if the rotational frequency of the axis (bearing) to be monitored changes.
The band limits are always adapted to the current rotational speed by the factor.
The calculation for this is Factor x rotational frequency = damage frequency range.
The factors are determined by the bearing manufacturers and can be obtained from them.
The speed must be provided to the sensor via one of the three paths.
- Pin 2 Input via a clock signal from an external sensor
- Process data output
- Static input of the parameter data
4.3. Absolute band limits
The absolute band limits can be used if the harmful frequency is known. For example, for electric motors where the frequency of the mains voltage is 50Hz (60Hz USA) and therefore does not change.
