Jügen Schuldt
3. EL CASO DE LAS COMUNIDADES CAMPESINAS
It is extremely difficult to perform synchronous time averaging on the signals measured from the bearing with a defective rolling element. This is due to the random phase of the impulses that are measured from such a bearing as shown in Figures 4.8(c), 4.9(c), and 4.10(c). As a result, the position of impulses in one time trace will be different than in the other traces, and when many averages from these traces are added together all the impulses that are characteristics of a defective bearing will disappear. These impulses are disappearing together with the random noise because of their randomness in phase. A new method called correlated time averaging is developed in this study to overcome this problem. One of the advantages from using this method is that a higher signal-to-noise ratio can be achieved even if the signals are measured without a trigger mechanism.
The main objective of the correlated time averaging method is to capture the family of impulses that occur due to the presence of a defect in a bearing component. In a rotating machine component such as bearings or gears, a defective component in operation will generate a family (or a series) of impulses. In each rotation, this family of impulses will repeat itself but the time of occurrence might be random depending on the type of defect present.
The first step in this method is to capture the family of impulses from a time trace. This is performed using a time window that slides along the time trace and capturing a family of impulses inside this window by positioning the window in the middle of a peak. Once the desired signal is captured it is saved for the next step and this procedure is repeated until a maximum number of captured signals is obtained for the next step of the averaging process. The procedure in these steps is shown in graphical form in Figure 4.17. The width of the capturing window is adjusted to optimise the total number of data points captured, so that enough data is obtained to represent the different types of defect that can occur. In this study, the total number of data points for the sliding window is set to 256. This represents 0.52 to 0.91 rotations of the bearing shaft depending on its speed. For the type of defects studied, this setting is enough to capture
the impulses and the different characteristics of each type of defects being studied can be identified.
Sliding window
Ampl.
t (sec)
Figure 4.17 Procedure for capturing a family of impulses from a time trace.
In the second step, a pair of the captured signals is selected and a cross correlation routine is performed. Then, these two signals are aligned so that the maximum coefficient of the cross correlation function is set to zero time. Next, the two captured signals are added and averaged to obtained an averaged time trace signal. Then, another pair of the captured signal is selected and this routine is repeated until there is no more signal left to be processed. Figure 4.18 below shows a schematic diagram of this procedure to obtain correlated time averaged signal from eight samples of time trace.
Captured Signals 1st Stage YSI'ft CA 2nd Stage ~(sr\) SZ CA s z S12 CA S 4 S31 CA S13 CA Stage S 3 CA S22 S14 CA S 3 Cross-correlation Averaging Routine
Figure 4.18 Schematic diagram on the correlated time averaging process from eight captured time traces.
Figure 4.19 shows some examples on the effectiveness of the correlated time averaging procedure to obtain a high signal-to-noise ratio from a defective rolling element bearing. As the number of averaging process is increased, the quality of the signal becomes better as shown in the figure. Comparison results between time-averaged signals obtained from a bearing with rolling element defect and time-averaged signals from a normal bearing are presented in Figure 4.20. As expected these signals represent amplitude modulation signals, whereby the higher frequency acts as a carrier frequency and the low modulation frequency contains the information that is characteristic of the bearing being tested.
The amount of noise reduced during the averaging process is dependent on the total number of averages used. Figure 4.21 shows a time trace of the noise reduced from thirty-two averaged signals. This signal is obtained by subtracting signal in Figure 4.19(d) from the signal in Figure 4.19(a). The percentage of noise reduced is calculated using the following formula:
Y * 2 _ V X 2
% of noise reduction = T"1 mycc/ x 100 z X „
(4.4)
where ximt is the instantaneous data, and xmged is the time-averaged data. The percentage of noise reduction from Figure 4.21 is calculated to be:
% of noise reduction = ( (1 5 2 8 4 .1 7 - 4 5 6 1 .2 9 6 ) / 1 5 2 8 4 .1 7 ) * 1 0 0
= 7 0 .2 %
This result shows that a large amount of noise is eliminated from the correlated time averaging process. Therefore, the presence of abnormality in bearing component is easily detected using this method as proven in this case study.