A function for counting cycles according to the Rainflow Method.
The cycle counting method implemented here is based on Downing's Method 1 defined in the paper "Simple rainflow counting algorithms" by S.D.Downing & D.F.Socie, International Journal of Fatigue, January 1982.
This method is for limited histories where the data has been obtained and stored. Method 2 from the same paper (not implemented here) is designed for open-ended histories, typically used in a monitoring situation.
numba is used to speed up the actual rainflow counting algorithm.
numpy is used for the data manipulation required for the method.
The result is quite a fast implementation.
import numpy as np
import rainflow
# .... retrieve stored time series data as a 1D numpy array.
ranges, means = rainflow.get_ranges_and_means(data)Sometimes it makes sense to convert real data into integral data in order to eliminate very small ranges caused by low-level noise.
Unless the data units are already such that 1 unit is a sensible rounding limit, the data will have to be scaled so that significant data is not lost.
import numpy as np
import rainflow
# .... retrieve stored time series data as a 1D numpy array.
scale = 10.0
data = rainflow.get_integral_data(data, scale) # data is now int32s.
ranges, means = rainflow.get_ranges_and_means(data)
ranges /= scale
means /= scale