Failure probability can be computed from the estimated survival probability over an observed time period. This probability is the residual after compounding survival probability over sections. This failure probability can be mapped to Time-to-Failure estimates using a sectional exponential decay model. This exponential decay curve is parameterized for accuracy, and varies slightly from one fault model to another. The more sections, the better the accuracy.
In the IT world, a difference of half a day in filing trouble tickets does not make a difference, so fewer sections are used. For other types of digital twins, finer sections are required for accuracy. A digital twin is designed as an accurate, virtual model of a physical device or object. Data scientists and IT personnel use digital twins to run simulations before committing to the time and cost of building physical devices or objects.
The following parameters can be tuned to adjust the Time-to-Failure:
A = 2000000
B = 14.72
B1 = 15.30
cutoff = 0.90
The following is an example python function:
def t2f_func(a, b, fp):
t2f = a * np.exp(-b*fp)
return t2f
failure_probability = 1 - np.prod(survival_probability_adjusted)
print ('Failure Probability: ', failure_probability)
if failure_probability >= cutoff:
print ('Time-to-Failure: ', np.ceil (t2f_func (A, B , failure_probability)), 'Days')
else:
print ('Time-to-Failure: ', np.ceil (t2f_func (A, B1, failure_probability)), 'Days')
Failure Probability: 0.9499373994767666
Time-to-Failure: 2.0 Days