Environmental Engineering Reference
In-Depth Information
• format of reliability data: operating time to failure or interval values
of intensities of the flow of failures;
• an assumption about the type of reconstruction of equipment after
failure and preventive maintenance: full recovery, partial recovery or no
recovery.
The following non-parametric methods for checking the presence of a
trend have been proposed and studied in Ref. 35:
• Laplace criterion (operating time to failure);
• test of two cells (operating time to failure);
• inversion method (the intensity of the flow of failures).
To test assumptions about the type of recovery equipment, the authors
of Ref. 38 proposed to use the Lewis-Robinson test.
Graphical analysis and non-parametric methods for testing the trend
are easy to implement and do not require special software, and the
accuracy of the results is inferior to more powerful parametric methods
of analysis. When the original data are accurate, these methods allow us
to quickly evaluate the primary data and give information about important
attributes such as the homogeneity of the sample, failure in running-in,
the level of uncertainty and importance of estimates. These methods are
well suited for preliminary data analysis and selection of components for
further analysis.
Requirements for reliability models are formulated depending on the
structure and volume of PSA and the available (or accessible) data on
reliability
39
.
The overall logic of the process shown in Fig. 2.10.
In Refs. 34, 35, 25, 40-42 the authors proposed and investigated cases
of generalised linear models to account for the reliability of ageing safety
equipment in PSA:
• linear model for the failure rate:
λ (
t
) =
a
+
bt
,
[2.11]
• Weibull model:
λ (
t
) =
at
b
,
[2.12]
• log-linear model:
λ (
t
) =
a
exp (
bt
),
[2.13]
Here λ (
t
) is the failure rate,
t
is time (age) of equipment,
a, b
are
parameters of the model, where
b
is a parameter characterising the presence
or absence of ageing. If
b
> 0, the failure rate increases with time,
b
= 0 -
corresponds to a constant failure rate and if
b
< 0 the failure rate decreases
with time (running-in failure).
Search WWH ::
Custom Search