Hello guys. I've been working on GFRP composites and characterization damage modes in composites with wavelet transform and ML algorithm. I was surprised by a strange thing. By use of continuum wavelet, my signal shows that three clusters of frequencies (0-160 KHz, 160-310 KHz and 310-520 KHz) are happening with each other at same time during the test but my cumulative energy-time plot show that by passing the time the slop of plot increases suddenly which shows that different types of damage happen respectively. Here are the pictures of another paper which had same issue.

Can any one explain for me what is the reason of this difference between wavelet transform and cumulative energy-time plot?