PSI - Issue 14

941 5

Vamsi Inturi et al. / Procedia Structural Integrity 14 (2019) 937–944 Vamsi, Sabareesh, Vaibhav/ Structural Integrity Procedia 00 (2018) 000–000

(a) Root crack - 50% Faulty

(b) Root crack - 100% Faulty

(c) Tooth chip - 25% Faulty

(d) Tooth chip - 100% Faulty

Fig. 3. Types of gear faults and their severity levels (in %)

Table 2. Dimensions of the faults

Dimensions of the fault (length * depth * thickness)

S.No.

Name/Condition of the gear

1 2 3 4 5

Healthy (H)

----

50% Faulty (F1) 100% Faulty (F2) 25% Faulty (F3) 100% Faulty (F4)

30 mm * 1.6 mm * 0.25 mm 30 mm * 4 mm * 0.25 mm

Tooth root crack

NA NA

Tooth chip

3. Feature extraction through EMD EMD is a non-linear multi-resolution self-adaptive technique capable of decomposing a complex signal into a set of complete IMFs without prior knowledge of the nature and number of IMFs present in the given signal [14 & 15]. IMFs represent the different fundamental oscillatory modes embedded in the signal. Each IMF has to satisfy the following criterion: (i) with in the whole data set, the number of extrema and number of zero crossings must either be equal or differ at most by one. (ii) at any point, the mean value of the envelope defined by local maxima and the envelope defined by local minima should be zero [14]. At the end of EMD process, a signal is decomposed into number of IMFs c i (i = 1, 2 ,3….I) and a residual r I . Original signal can be obtained by summing all the IMFs and the residual as shown in Eq. (1). (t) = ∑ � � ��� + � (1) The IMFs c 1 , c 2 …… c I represent various frequency bands ranging from low to high. The frequency components present in a particular frequency band are different and their value changes with the variation in the original signal x (t) [16]. Noise presented in the data signal may have decisive influence on the decision making. Therefore, it is important to make sure that the input signal is noise free while performing the EMD process. At the very first step, the noise present in the raw data signal is filtered out and the filtered signal is then sent as an input to EMD. A moving-average filter of window size 6 is used to accomplish the smoothing action on to the acquired raw signal which in turn reduces the amount of random noise which is present in the signal. In the next step, EMD decomposes the signal into a finite number of IMFs. The following procedure is followed for performing the EMD process on the recorded experimental data. (1) Original signal x (t) containing data for the 0.5ms time period is taken as the input and it is plotted in the time domain to determine the maxima and minima. (2) All the maxima are joined together using cubic spline curve producing an upper envelope. Similarly, lower envelope is produced by joining all the minima. (3) Average value for each data point of upper and lower envelope is determined. After this, average is subtracted from the original signal to obtain a new signal, x 1 (t).