Citation: An, D. Prediction-Interval-
Based Credibility Criteria of
Prognostics Results for Practical Use.
Processes 2022, 10, 473. https://
doi.org/10.3390/pr10030473
Academic Editor: Bhavik Bakshi
Received: 21 January 2022
Accepted: 24 February 2022
Published: 26 February 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the author.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Technical Note
Prediction-Interval-Based Credibility Criteria of Prognostics
Results for Practical Use
Dawn An
Advanced Mechatronics R&D Group, Daegyeong Division, Korea Institute of Industrial Technology,
Daegu 42994, Korea; dawnan@kitech.re.kr
Abstract:
Prognostics is an AI-based technique for predicting the degrading/damaging behavior and
remaining useful life (RUL) of a system, which facilitates a cost-effective and smart maintenance
process. Many prognostics methods have been developed for various applications, such as bearings,
aircraft engines, batteries, and fuel cell stacks. Once a new prognostics method is developed, it is
evaluated using several metrics based on the true value of the RUL. However, these typical evaluation
metrics are not applicable in real-world applications, as the true RUL cannot be known before the
actual failure of a system. There are no ways to determine the reliability of prognostics results
in practice. Therefore, this article presents the credibility criteria of prognostics results based on
prediction intervals (PI), which are known values, unlike the true RUL. The PI-based credibility
criteria for prognostics results are explained with two simple examples under different levels of noise
to help with the decision making on prognostics results in the industrial field.
Keywords:
prognostics; remaining useful life; metrics; credibility criteria; prediction interval;
maintenance
1. Introduction
Prognostics is one of the disciplines in condition-based maintenance [
1
], which is a
cost-effective maintenance strategy that repairs damaged parts only when required. The
purpose of prognostics is to predict the remaining useful life (RUL) of the system, which is
defined as the remaining cycles/time before maintenance and has been studied for the past
few decades in many engineering applications, such as bearings [
2
,
3
], aircraft engines [
4
],
batteries [
5
], and fuel cell stacks [
6
]. RUL can be predicted using degradation data obtained
up to the current time and several prediction algorithms, such as particle filters [
7
,
8
] and
artificial neural networks [9,10].
The results of RUL prediction are updated whenever new measurement data are
available during the life span of a system; an instance of such a system is illustrated in
Figure 1. In Figure 1a, pentagrams, squares, and circles represent the median, lower bound,
and upper bound of the RUL prediction results, respectively. We assume that this system
has been operating for two cycles as of now. In this case, the median RUL (the pentagram
at two cycles in Figure 1a) indicates that this system needs maintenance approximately one
cycle later. Thereafter, this system is operated for one more cycle, and the current duration
becomes three cycles. It is found that maintenance is not necessary, and the median of
RUL prediction is even increased to approximately one and a half cycles. The RUL is the
remaining number of cycles and should thus decrease as the operating cycles of the system
increases; however, the prediction results show an increasing tendency of up to five cycles.
Although RUL prediction results decrease after one cycle, that is, the current duration is six
cycles, there is no way to determine the reliability of the current prediction result by only
knowing the past prediction results (note that the RUL results after the current six cycles
are unknown as of now). A typical way to evaluate the performance of RUL prediction is
to employ the true value of the RUL, as shown in Figure 1b. The solid diagonal line is the
true RUL, and it is now clear whether the prediction results are reliable.
Processes 2022, 10, 473. https://doi.org/10.3390/pr10030473 https://www.mdpi.com/journal/processes
