172
S. Wu et al.
continuous accumulation of damage, or sudden fracture of components may even
lead to catastrophic accidents. Therefore, it is extremely important to prevent fatigue
damage of metal materials in service at an early stage. The characteristic parameters of the acoustic emission signal include ringing count, amplitude, RMS voltage,
energy. Especially energy is widely used as a characteristic parameter. Here, these
characteristic parameters are generalized to the characteristic parameter analysis of
MAE signal. However, if the characteristic parameters of MAE signals interfered
by strong background noise are directly extracted, the results obtained are not ideal
or even contrary to the actual situation. Therefore, effective new methods must be
sought to achieve the noise reduction function of MAE signal.
For non-stationary and nonlinear signals, the literature [1] presents a CEEMD
method, which converts the white noise added in the EMMD algorithm into positive
and negative pairs. It can better eliminate the residual noise of the reconstructed signal
and improve the signal-to-noise ratio (SNR) of the signal. In the field of device
diagnostics, the CEEMD algorithm is applied to the fault detection of induction
motors [2], and the fault of the motor is detected by the method of spectrum analysis
of the IMF component. In the field of geophysics, the CEEMD algorithm is applied to
denoising seismic radar signals [3, 4], which successfully eliminates coherent noise in
seismic radar signals and improves deeper response. CEEMD has been successfully
applied to biomedical [5, 6] mechanical fault diagnosis [7–9], but its research in the
field of MAE is relatively less. Based on the unique advantages of CEEMD and the
characteristics of MAE signals with weak, low amplitude and easy to be submerged
by noise, the CEEMD method is introduced into the characteristics extraction of
MAE signals. A method for extracting characteristic parameters of MAE signals
based on CEEMD is proposed, and experimental verification is performed under
fatigue condition, which shows the unique advantages of this method.
16.2 Method and Experimental
Huang [10] first proposed a time-frequency analysis method of Empirical Mode
Decomposition (EMD), which is an adaptive signal processing method. It decomposes the signal into a finite IMF, which can accurately express the different frequency
components contained in the signal. Therefore, EMD is a time-frequency analysis
method for nonlinear steady-state signals. It can repeatedly subtract the envelope
mean to eliminate the oscillation and effectively reflect the transient characteristics
of the signal. For a given MAE signal x(t), firstly, select all the extreme points on x(t)
and connect the maxima with the cubic spline to form the upper and lower envelopes.
Then take the median value m 1 (t) for the upper and lower envelopes, take the temporary local oscillation h 1 (t) = x(t) − m 1 (t), replace x(t) with h 1 (t), repeat the above
two steps until m(t) is equal to 0, the first IMF component is c 1 (t) = h 1m (t), the
residual r 1 (t) = x(t) − c 1 (t), and x(t) is replaced by r 1 (t), repeat the above four
steps until the decomposition cannot continue. There is a sum of n IMF components
and a residual term:
S. Wu et al.
continuous accumulation of damage, or sudden fracture of components may even
lead to catastrophic accidents. Therefore, it is extremely important to prevent fatigue
damage of metal materials in service at an early stage. The characteristic parameters of the acoustic emission signal include ringing count, amplitude, RMS voltage,
energy. Especially energy is widely used as a characteristic parameter. Here, these
characteristic parameters are generalized to the characteristic parameter analysis of
MAE signal. However, if the characteristic parameters of MAE signals interfered
by strong background noise are directly extracted, the results obtained are not ideal
or even contrary to the actual situation. Therefore, effective new methods must be
sought to achieve the noise reduction function of MAE signal.
For non-stationary and nonlinear signals, the literature [1] presents a CEEMD
method, which converts the white noise added in the EMMD algorithm into positive
and negative pairs. It can better eliminate the residual noise of the reconstructed signal
and improve the signal-to-noise ratio (SNR) of the signal. In the field of device
diagnostics, the CEEMD algorithm is applied to the fault detection of induction
motors [2], and the fault of the motor is detected by the method of spectrum analysis
of the IMF component. In the field of geophysics, the CEEMD algorithm is applied to
denoising seismic radar signals [3, 4], which successfully eliminates coherent noise in
seismic radar signals and improves deeper response. CEEMD has been successfully
applied to biomedical [5, 6] mechanical fault diagnosis [7–9], but its research in the
field of MAE is relatively less. Based on the unique advantages of CEEMD and the
characteristics of MAE signals with weak, low amplitude and easy to be submerged
by noise, the CEEMD method is introduced into the characteristics extraction of
MAE signals. A method for extracting characteristic parameters of MAE signals
based on CEEMD is proposed, and experimental verification is performed under
fatigue condition, which shows the unique advantages of this method.
16.2 Method and Experimental
Huang [10] first proposed a time-frequency analysis method of Empirical Mode
Decomposition (EMD), which is an adaptive signal processing method. It decomposes the signal into a finite IMF, which can accurately express the different frequency
components contained in the signal. Therefore, EMD is a time-frequency analysis
method for nonlinear steady-state signals. It can repeatedly subtract the envelope
mean to eliminate the oscillation and effectively reflect the transient characteristics
of the signal. For a given MAE signal x(t), firstly, select all the extreme points on x(t)
and connect the maxima with the cubic spline to form the upper and lower envelopes.
Then take the median value m 1 (t) for the upper and lower envelopes, take the temporary local oscillation h 1 (t) = x(t) − m 1 (t), replace x(t) with h 1 (t), repeat the above
two steps until m(t) is equal to 0, the first IMF component is c 1 (t) = h 1m (t), the
residual r 1 (t) = x(t) − c 1 (t), and x(t) is replaced by r 1 (t), repeat the above four
steps until the decomposition cannot continue. There is a sum of n IMF components
and a residual term: