You are on page 1of 17

Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

https://doi.org/10.1007/s40430-021-03222-y

TECHNICAL PAPER

Asset management strategies using reliability, availability,


and maintainability (RAM) analysis
Maria Valentina Clavijo Mesa1   · Carmen Elena Patino‑Rodriguez2 · Fernando Jesus Guevara Carazas3 ·
Indra Gunawan4   · Enrique López Droguett5 

Received: 18 May 2021 / Accepted: 27 September 2021 / Published online: 12 October 2021
© The Brazilian Society of Mechanical Sciences and Engineering 2021

Abstract
Reliability, availability and maintainability (RAM) analysis is used for various strategic purposes in the operations of
productive assets to identify key points to apply improvements to. The conventional RAM approaches are, nevertheless,
practical guides developed for case studies, not for analyses that consider any level of complexity asset. This paper presents
a methodology for RAM analysis, which represents a logical pathway for analysts and operators of engineered systems
to guide the decision-making of asset managers. For different scenarios, strategies to collect the required information are
discussed as well as recommendations to make the asset more efficient. Due to the usual limitation of available resources
in the companies, and prioritization maintenance actions that must be assertive to achieve the best possible efficiency dur-
ing the operation, the methodology includes techniques to identify critical equipment as structural importance measures
(Birnbaum, Risk Reduction Worth, Risk Achievement Worth). The proposed methodology is demonstrated by developing a
RAM analysis for a heavy-duty truck. This type of vehicle is used for collecting cargo in a Colombian city. The case study
results in operational, tactical, and strategic recommendations for operation and maintenance management professionals.

Keywords  RAM analysis · Importance measures · MTBF · MTTR​ · Fitted distributions · Heavy-duty trucks

1 Introduction need for quality at the best possible cost is a requirement


for companies to remain current in a competitive environ-
The industrial production process has evolved with import ment. In this context, RAM analysis has been presented as
changes in the quality of product, operations management, a powerful tool to estimate a set of indicators that allow
and recent technology. According to Tsarouhas [58], the monitoring asset health management.
Nevertheless, technological advances and improvements
have caused increasing system complexity. Concurrently,
Technical Editor: Andre T. Beck.

2
* Maria Valentina Clavijo Mesa Analytics and Research for Decision‑Making Laboratory
valentina.clavijo@usp.br (ALIADO), Industrial Engineering Department, University
of Antioquia, Calle 67 # 53‑108, Medellín, Colombia
Carmen Elena Patino‑Rodriguez
3
elena.patino@udea.edu.co Asset Management, Operation and Maintenance Research
Group (GOMAC), Mechanical Engineering Department,
Fernando Jesus Guevara Carazas
National University of Colombia, Carrera 64C #63‑120,
fjguevarac@unal.edu.co
Medellín, Colombia
Indra Gunawan 4
Adelaide Business School, University of Adelaide, Adelaide
indra.gunawan@adelaide.edu.au
SA 5005, Adelaide, SA, Australia
Enrique López Droguett 5
Civil and Environmental Engineering Department
eald@ucla.edu
and Garrick Institute for the Risk Sciences, University
1 of California, Los Angeles, 404 Westwood Plaza,
Analysis, Evaluation and Risk Management Laboratory
Engineering VI, 580B, Los Angeles, CA, USA
(LabRisco), Naval Architecture and Ocean Engineering
Department, University of São Paulo, Av. Prof. Mello
Moraes, 2231, Cidade Universitária, São Paulo, Brazil

13
Vol.:(0123456789)
495 
Page 2 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

requirements for high functionality and high performance in the red color for the concepts that were developed first
also increase. These conditions result in challenges and (1983) and ends in the yellow color to represent the latest
opportunities [13]. For example, component reliability is concepts developed in this topic (2021).
often nonspecific since reliability depends on the operating The first RAM concepts were associated with Maintain-
and environmental conditions [39]. Maintainability, in turn, ability analysis, then Reliability analysis was developed;
is difficult to estimate due to the uncertainty associated with finally, Availability analysis emerged. In the decades of
repair processes in complex assets, and estimating the availa- comprehending the 2000s, it seems that the main concern
bility of a system tends to require a high computational cost. regarded the solution methods, cost, and performance. For
In this sense, Sharma et al. [50] analyzed the literature about instance, Liberopoulos and Tsarouhas [34] developed a sta-
maintenance optimization models, concluding that mainte- tistical analysis of failure data of a production line. For the
nance models require maintenance teams to have software study, the line failures were collected for four years. The
for information storage and subsequent efficient retrieval for analysis included the computation of the descriptive statis-
maintenance management. tics of the failure data, the characterization of the relevant
To obtain a rigorous and real summary of the RAM anal- failures, the computation of the parameters of the theoretical
ysis, bibliometric analysis was performed to quantitatively distributions that best fit the failure data, and the investiga-
evaluate the eligible articles and identify quantifiable pat- tion of the existence of autocorrelations and cross-correla-
terns of interest, such as the occurrence of keywords distri- tion in the failure data.
bution in articles. Descriptive information, including title, More recently, the concern has revolved around an inte-
keywords, journal name, abstract, and publication year were grated RAM analysis and engineering systems applications
derived. Scopus database and VOSViewer [61] were used mixing advanced computational concepts and risk analy-
collectively to compile, manage, and analyze the descrip- sis, such as Gao and Wang [21] and Kumar and Jain [30],
tive properties of the eligible articles. The network for key- who used Markovian approaches. Kumar and Jain [30]
words of RAM analysis publications is presented in Fig. 1. investigated the reliability of a multi-component machin-
This methodology is based on keywords used in academic ing system with unreliable servers. In the analysis, transient
research publications that, in our case, are related to the probabilities are obtained by computing eigenvalues of the
RAM concepts and techniques used to estimate each of the transition matrix and using a spectral method. Particularly in
RAM indicators. this paper, the metrics are estimated to be great for numeri-
A closer look at Fig. 1 shows the temporal evolutions cal experiments. Gao and Wang [21] expose a reliability
associated with the RAM concepts. The color scale begins and availability analysis of a repairable retrial system with

Fig. 1  Network of RAM analy-


sis publications

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 3 of 17  495

mixed standbys. Solving equations of statistical equilibrium of time between failure, and Time To Repair (TTR) data and
in matrix form, the authors obtain the asymptotic availability their parameters. Sikos and Klemeš [53] design a fault tree
of the machine system as well as the expected busy cycle (FT), to analyze the reliability of heat exchanger networks,
of the repair facility. Furthermore, Sharma and Kumar [51] and Martorell et al. [36] suggest the use of genetic algorithms
developed a RAM analysis in a urea plant, situated in north- to estimate the availability of industrial installations.
ern India. During this application, the authors conducted the In another study, Cheung and Wang [9] quantified the
analysis applying a Markov model. reliability and availability of a cooling system. The study
Yang and Tsao [63] present an alternative to estimate also defines which types of cooling system designs are
the Mean Time To Failure (MTTF) based on the Laplace preferable for highly reliable centers. The exponential
transform technique. In the paper, the authors focus on distribution is used to model the failure data (assuming
reliability-based measures of standby systems with retrials constant hazard rate) and the asymptotic availability of
and working vacations. For this reason, it is simulated as data center cooling equipment from the field and calcu-
a repairable system consisting of M primary components, lating the availability of complete systems with different
S spare components and a repairman. The repair times of numbers of redundant equipment and distribution headers.
failed components are assumed as exponential. Other applications of RAM analysis in the industry
We would like to point out that several distributions have are presented by Arias Velásquez and Mejía Lara [5] and
been adopted in the RAM analysis. For example, Garg [22] Herder et al. [26]. In the first, using Monte Carlo simu-
during the estimation of RAM indicators, the author quanti- lation (MCS) and reliability block diagrams (RBD), the
fies the uncertainty in the data for a crankcase manufacturing authors describe the implementation of a RAM simula-
plant. The author applies the particle swarm optimization tion model for a plastic plant located in the Netherlands.
(PSO) technique, and for uncertainty analysis, triangular and Arias Velásquez and Mejía Lara [5] developed a RAM
normal distribution using a setting suggested by: (i) deci- study for series capacitor banks for power lines located at
sion-makers, (ii) design maintenance experts, and iii) system substations in Peru. In that research, it was assumed that
reliability analysts. PSO technique is also implemented by the exponential distribution was adequate for modeling
Chalabi et al. [8] to study an optimization problem of the the failure rate of the electrical equipment, and MCS was
grouping strategy of preventive maintenance tasks for multi- also used to estimate the annual availability and the MTBF
unit series production systems. (mean time between failures) of the system.
In short, even when the RAM concepts were known a However, these RAM approaches suffer limitations as
few decades ago, these concepts continue to be of interest they fail to:
to the scientific community, which has related the concepts
to some other areas. In fact, as evident from the above stud- –  Integrate each RAM indicator with the general analysis
ies, systems models are developed to solve reliability, avail- of the asset, concluding about alternatives for the asset
ability, and maintainability problems. However, Zhang et al. manager that allow improving the reliability, availabil-
[64] argue that there are still unresolved challenges for RAM ity and/or maintainability of their asset,
analysts, for example: –  provide strategies that researchers could apply during
the data collection required to RAM analyses,
• Familiarization with the engineering system under study –  present a methodological graph of the process imple-
depends on the experience and competence of RAM mented during the analysis, instead of a figure of the
engineers instead of designers, thus creating the possi- technique used,
bility that key aspects of the engineering of the system –  expose analysis techniques that could be implemented
may be ignored or misinterpreted. in computational tools for free access and low compu-
• The interactions between components and functions are tational cost
not sufficient to evaluate RAM performance because
the failure effect is only identified and evaluated in the From these premises and considering that during the stud-
selected hierarchical decomposition, and health is evalu- ies presented there was no evidence of an integrated meth-
ated in a remarkably similar way. odology for RAM analysis that can be applied to systems
of different levels of complexity. The authors of this paper,
Then, to explore the potential of RAM analyses, it is relevant hereinafter called the authors, were motivated to design a
to cite some studies, such as Choudhary et al. [11], Mar- RAM analysis methodology to support the four limitations
torell et al. [36], and Sikos and Klemeš [53]. Choudhary et al. previously mentioned. Additionally, the proposed meth-
[11] implement RAM analysis of a cement plant in Northern odology is demonstrated through a case study involving a
India, using operational data collected for two years. Subse- heavy-duty truck (HDT) used for a Colombian company
quently, the authors used the tests of lack of fit, distribution that collects cargo in Colombian cities.

13
495 
Page 4 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

We here proposed a methodology for RAM analysis and represents the theoretical base for RAM analysis, and nor-
strategies to collect the required data for analysis are detailed mally, the process developed is presented in a specific con-
(Sect. 2). Afterward, the application of the methodology is figuration of the asset (diagram functional description).
elucidated, using a HDV as a case study (Sect. 3). Lastly, in
Sect. 4, the concluding statements are presented.
2.2 Reliability analysis

2 Methodology Secondly, the methodology proposes executing the reliabil-


ity analysis. For this analysis, the inputs required are the
The proposed methodology is based on traditionally applied mission time and hazard rate of the system components. The
concepts. Yet it is innovative because it presents a clear and mission time is the period necessary to calculate the prob-
replicable methodological structure so that readers and ana- ability of non-failure [39]. The hazard rate could be defined
lysts of the health status of productive assets can replace as the frequency at which an engineered component fails,
the study and thus monitor the RAM indicators required for expressed in failures per unit of time [20]. Figure 3 presents
the proper management of their assets, besides integrating some strategies to collect information related to hazard rates
several techniques to evaluate the reliability, maintainability, of components.
and availability of an asset, generating different alternatives During reliability analysis, the analyst should select the
for the analyst according to the system complexity. Figure 2 analytical logic techniques to quantify the probability of the
shows the methodology diagram. system to perform without failure under a given set of oper-
The methodology is divided into four main steps: func- ating conditions for a mission time, given it was operating
tional description of the system, reliability analysis, main- at the initial instant. Thus, Fig. 2 presents some of the mul-
tainability analysis, and availability analysis. The first step tiple alternatives that might be applicable, such as fault tree
consists in defining the technical and operational charac- analysis (FTA), RBD, Markov Chain, or Bayesian networks
teristics of the system. As mentioned by Stapelberg [56], (BNs) [6, 19, 40]. Particularly, FTA and RBD are widely
engineering integrity includes RAM of inherent systems used formalisms in system reliability modeling [18].
functions and their related equipment. Thus, the analyst Table 1 summarizes some reliability techniques and pre-
establishes the characteristics that define the engineering sents the limitations that must be previously considered by
principles and the operational characteristics of the equip- analysts.
ment, from information from manufacturers or other sources Note that FTA and RBD have two different approaches,
of technical information. that is, in the RBD, the asset (productive system) is repre-
sented by functionally interconnected components or reli-
2.1 Functional analysis ability relationships, whereas in the FTA, it shows which
combinations of failures between components trigger in a
Functional analysis should be the analyst’s first approach to total or partial failure of the entire asset or system [62]. In
the asset under study [48]. Particularly, Fig. 2 exposes three this sense, the recommendation for the analyst is that FTA
alternatives as possible sources of information to functional implementation could be more advantageous in case the
analysis. Generally, the information is obtained for com- hazard rate is collected for events (for example, piston pres-
ponents; therefore, the analyst must collect as much infor- sure too high, cylinder excessive external load, ring pressure
mation as possible from the components and later design too low, etc.); whereas application of RBD may be more
the functional analysis of the system. Thus, manufacturer’s appropriate if the hazard rate is compiled for components
manuals and literature reviews could be used for designing (for example, loss of pump, valve fail, switch failed.) and
system functional trees. A functional tree is a graphical tool the functionality of the system can be decomposed into
that allows establishing the dependencies or functional links series–parallel systems.
between the subsystems of an entire system. Basically, in The results expected from the reliability analysis are the
functional analysis, these trees show the existing relations estimation of asset failure probability for predefined mission
within a system by decomposing a problem into simpler time (unreliability) and the identification of the most critical
parts or elements [24]. Note that the failure mode and effect components of the asset. However, some authors propose
analysis (FMEA) guidelines are related to the industry to that system critical components can be established by the
which the asset being analyzed belongs [1, 41, 47, 59]. qualitative evaluation of FTA. The methodology proposed
Once the functional analysis and FMEA is implemented, in this article suggests using the importance measures (IM)
the analyst is expected to identify the equipment bounda- for quantitatively assessing the most critical components.
ries, assumptions of asset operation, and redundancy levels Table 2 presents a summary of information about structural
of the system. Therefore, the first step in the methodology IM commonly used.

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 5 of 17  495

Fig. 2  Proposed methodology

13
495 
Page 6 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

Fig. 3  Data collection strategies

In Table 2, the following definitions are used: Table 3 summarizes some maintainability techniques
and presents the limitations that must be previously con-
Rs the present system reliability. sidered by analysts.
Ri (the present
) reliability of component i . Upon obtaining the repair times of the system, the
Rs Ri = 1 the increased reliability level with component methodology proposes plotting histograms for these times
i optimized) or assumed to be perfectly reliable. to know the data frequency distribution. Subsequently, the
Rs Ri = 0 the decreased reliability level without compo- analyst should apply parametric or non-parametric meth-
(

nent i or with component i assumed failed. ods to estimate the theoretical or empirical probability
density function (pdf) of the repair time system. Some
To explore more information about IM, the reader is encour- references recommended to estimate the theoretical or
aged to consult authors such as Clavijo Mesa [12], Modarres empirical pdf are Chouaib [10], Clavijo Mesa [12], and
[38], Si et al. [52], and Van Der Borst [60]. Silverman [54].
The focus of the maintainability analysis is to appraise
the probability to repair the asset in time t and evaluate if
2.3 Maintainability analysis there is a suitable probability distribution family to model
the asset repair times.
According to the methodology, the next step encompasses
the maintainability analysis. Providing information about
the repair rate of components and mission time of system/ 2.4 Availability analysis
asset is required. Figure 3 presents some strategies to collect
information related to the repair rate of components. Regarding availability analysis, the methodology suggests
Once the input information is compiled, the analyst the use of the systems MTBF and Mean Time To Repair
defines the method for emulating the system operation and (MTTR). Note that using the simulated times in maintain-
maintenance. Some alternatives are MCS, Markov Chains, ability analysis, it is possible to estimate these times. At
Dynamic FTA, Dynamic RBD, Dynamic Bayesian Network this point, the analyst can select between different availabil-
(DBN), or Petri Nets [16, 46]. Note that traditional reliability ity approaches, that is, from MTBF and MTTR, the analyst
techniques of analysis, such as FTA and RBD, are unable to can estimate the steady-state availability and/or inherent
model the dynamic behavior of systems (as transitions of availability.
unavailable state to available state due to process of failure- Note that the steady-state availability, A(∞) , is the
repair) [28]. For this reason, the implementation of other limit of the availability function as time tends to infinity.
techniques is suggested in the methodology. The steady-state availability is also called the long-run or

13
Table 1  Some reliability techniques
FTA RBD Markov chain BN

Logic diagrams that display the state Graphical technique, which Stochastic process of parameters Directed acyclic graph I which each
of systems (top event) in terms of expressed the concerned system as whose development may be node represents a random variable,
the states of its basic events [31] connections of a number of com- conducted through a series of and the directed edges between vari-
ponents in accordance with their transitions between the states of a ables [33]
logical relation of reliability [25] system [43]
Limitations
Degraded performances cannot be ✓ ✓
considered
Priority and standby condition of ✓
components cannot be treated
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

Incapacity to model time interval ✓


too short
Common cause failure not modeled ✓
It is not very useful for explaining ✓
events
Conditional probability is required ✓
Huge computational burden ✓ ✓

13
Page 7 of 17  495
495 
Page 8 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

Table 2  Summary information about Birnbaum, RRW, and RAW Importance Measures


Measure Abbreviation Interpretation Principle

Birnbaum importance BI High values indicate that a small change in the reliability of component i, will Rs (Ri = 1) − Rs (Ri = 0)
result in a large change in the system reliability
Risk reduction worth RRW​ It presents the maximum increase in reliability for an improvement to the Rs ∕[Rs (Ri = 1)]
component i
Risk achievement worth RAW​ It presents the maximum decrease in reliability for a decadence of the com- [Rs (Ri = 1)]∕Rs
ponent i and indicates the importance of maintaining the current level of
reliability for the component i

asymptotic availability [7]. A common equation for the analyst can evaluate the quality of the data and verify if it is
steady-state availability found in the literature is: useful for the purposes of the analysis.
[ ] Normally, the data used in this strategy are (a) comput-
A(∞) = lim A(t) = lim e − ∫ ((t)+r (t)) dt
∫ (t)e∫ ((t)+r (t)) dt
dt + c erized database (typically grouped in data tables with sev-
t→∞ t→∞ r eral equipment attributes); (b) published data handbooks
(1) (sometimes simplified versions of computerized databases,
However, the case of analysis only considers the down- but the formats can depend on the publisher) and (c) expert
time due to corrective maintenance of the system [3]. The judgment (published data handbook or results from using
steady-state availability could be defined as the inherent statistical expert analysis methods) [27].
availability, If there is sufficient data volume, the analyst could imple-
ment an exploratory data analysis. However, in other cases,
A(t) = MTBF∕(MTBF + MTTR) (2)
some distribution in conformity with the component life
Finally, the results of the RAM analysis are reported, and cycle is assumed.
the analyst would know the effectiveness of the asset once In an observational study (see Fig. 3), specific data are
identifying how the asset fails, what the expected time for collected. The intention is to use the information to gain
the systems to return to operation is and, in the short or insights into the performance of assets under analysis. These
long term, what the likelihood is that the system will oper- data can be established by one company from: enterprise
ate satisfactorily. Thus, in this report, the analyst will not installations, and its key performance indicators (KPI).
only present the MTTF or MTTR of the asset as commonly Another alternative consists in applying Bayesian inference
occurs, but will also be able to expose a ranking of critical towards updating retrospective data as more evidence or
components or subsystems and even size the maintenance information becomes available about the failure and repair
team once the researcher knows the pdf of the system repair. process of assets [57].
Nevertheless, the methodology is flexible and, therefore, in Experimental studies produce data obtained because of
Fig. 2, the analyst is considered to have previously known active interventions. The analyst systematically changes or
some indicators (reliability, availability or maintainability) manipulates, a set of experimental factors while controlling
of the asset. For example, if the analyst knows the reliability for other variables. Particularly, these studies can apply to
of the asset and only owns the repair rate of components, new technology equipment for which experience data do not
then an A&M analysis could be developed, implementing yet exist. In this strategy, the analyst could implement relia-
the maintainability analysis, and availability analysis as pre- bility-based life testing or accelerated lifetime testing. Once
viously discussed. data are collected, the practitioner may test distributions of
It is worth highlighting that the results from applying the the estimated hazard rate and the repair rate of the asset.
methodology depend on the quality of the data used. Thus, To support the maintenance and operation decision-
Fig. 3 presents some strategies to collect the hazard rate and making in a Colombian company, the authors applied this
repair rate of system components. methodology to a heavy-duty truck (HDT). The analysis
In general terms, the asset manager could select between implementation and results are presented and discussed in
three strategies to collect failure and repair rates of asset the next section.
components: (a) retrospective study; (b) observational study,
or (c) experimental study.
Retrospective studies are based on historical data. These
data describe past process performance. Nevertheless, the

13
Table 3  Some maintainability techniques
MCS Markov chain DFTA DRBD DBN Petri nets

Generates random Consists a finite number Extends traditional FTA The condition of each The fundamental assump- Considers discrete events
numbers from random of states with the by defining 4 additional component is charac- tion in the case of time that are characterized as
variables for modeling Markovian property gates: Priority AND, terized by a variable series modeling is that being concurrent, syn-
risk or uncertainty of a and some transition sequence enforcing, state identifying the an event can cause chronous, asynchronous,
certain system [40] rates λij [32] standby and functional operational condition another event in the distributed, parallel,
dependency [19] of the component at a failure, but not vice nondeterministic, and
given name [18] versa [6] stochastic [16]
Limitations
Forces us to use too ✓
complex a model
Collecting and structur- ✓
ing expert knowledge is
a challenge
Incapacity to model time ✓
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

interval too short


Limitations in modular ✓ ✓
analysis
It is not very useful for ✓
explaining events
Approximate technique ✓
The concurrency of ✓
operations has become
more common increas-
ing the complexity
Huge computational ✓ ✓
burden

13
Page 9 of 17  495
495 
Page 10 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

3 Case study system is responsible for maximizing the friction between


the tires and the pavement to give steering stability and
A HDV is a vehicle designed to have a higher gross vehi- firm handling.
cle weight (GVW). HDTs are usually designed to have The steering system allows the driver to steer the vehi-
a GVW of 3500 kg or more [49]. This type of vehicle cle safely and precisely. The power transmission system is
is interesting because, according to Giakoumis [23], required for transferring engine power to the driveshaft and
HDT or just Heavy Trucks are not manufactured on large rear wheels; the gear inside the transmission changes the
scales. The design and manufacturing processes of this vehicle’s drive-wheel speed and torque in relation to the
type of vehicle are characterized by the great variety engine speed and torque. A truck transmission often operates
of engine—systems—chassis combinations that can be for periods exceeding a million and a half kilometer, with
obtained according to the specific need of operation or simple maintenance activities, such as checking lubricant
fleet design. levels and preventive changes.
This section details the application of the proposed meth- Regarding FMEA, the authors defined the analysis struc-
odology for analyzing HDTs used for a Colombian company ture using the FMEA handbook published by the Automotive
that collects cargo in Medellin city. Industry Action Group and Verband der Automobilindustrie
in 2012 [1]. Failure reports of the HDT fleet between 2015 and
3.1 Functional Analysis 2019 were also used in the Risk Priority Number (RPN) esti-
mation. Table 4 presents the most critical failure modes.
The function of HDT under consideration is transporting a Note that the power transmission system is critical
cargo of up to 12 tons from the urban area of the city to the because failures in flanks and crosses could generate detach-
company facility. This operation occurs at altitudes between ment of the Cardan axis. Additionally, failure modes associ-
3608 and 5905 ft. above sea level. Using the Manufactur- ated with brake pneumatic systems also present high RPN;
er’s Manual and a Literature Review, defined the functional this is explained by the potential safety negative effect of an
structure of the asset (see Fig. 4). eventual failure.
HDT consists of seven systems. The electrical and con-
trol system is responsible for supplying power to the starter 3.2 Reliability analysis
motor, besides providing electricity for the lighting and
security system, the instrument and meter system, as well As previously pointed out, reliability is the probability that
as for the computers operation. Additionally, this system a system will execute a designated function, in a defined
provides control regarding the position of the vehicle and period, under specific operating and maintenance conditions.
warns in the event of a critical component failure. Brakes To obtain this probability for HDT, the authors developed
and pneumatic systems correspond to a type of friction a retrospective study. Thus, the authors used a computer-
brake for vehicles in which compressed air pressing on ized database of failure data compiled during a time horizon
a piston is used to apply the pressure on the brake pad of approximately 47 months (December 2015–November
needed to stop the vehicle. The engine system oversees 2019).
converting heat from burning gas into mechanical energy. As there was sufficient data volume, the analysts imple-
Particularly, the HDT under study has an internal com- mented an exploratory data analysis to estimate the TTF
bustion engine with a configuration of six cylinders in a (hours) of each vehicle system, and then probability distribu-
straight line. The chassis responds to the basic vehicle-cab, tions (Exponential, Weibull, Lognormal and Gamma) were
frame and running gear, whereas the bodywork system is fitted for each of them. These probability functions must be
the container in which the load is carried. The suspension tested using Goodness of Fit (GOF) tests; particularly in this
study, Kolmogorov–Sminorv is used as a GOF test, which is

Fig. 4  Functional tree of heavy-duty truck

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 11 of 17  495

available in statistical software R. In case the reader wants assuming that systems are independent, the reliability of
to explore different GOF tests, Anderson and Darling [4], HDT, Rs (t) could be calculated as follows:
Csorgo and Faraway [15] and Marsaglia et al. [35] are some n
references that detail the test process.

Rs (t) = Ri (t) (3)
Table 5 presents the fitted parameters in each case and i=1
the p-value obtained from the Kolmogorov–Smirnov test
that supported that distribution. Likewise, the MTTF where Ri (t) is the reliability of the system i for time t  , and n
is presented, calculated as the mean of each of the fit- is the number of systems that compose the asset.
ted distributions. Note that the p-value is a measure of Again, using the free software environment for statisti-
the strength of the sampling evidence against the null cal analysis, R, the authors developed a code to quantify
hypothesis. A very low or less than 0.05 (significance the HDT reliability [44]. Figure 6 shows the results for
used) p-value will mean great sampling evidence against each system (a) and for the vehicle (b).
the established null hypothesis. Note that the electric and control system presents the low-
Using the hazard rate of each system and considering est reliability, whereas the engine system exposes the highest
that operators are interested in quantifying the HDT reli- reliability values over time. Regarding HDT, Fig. 6b reveals
ability for mission times of 250 operational hours, the that after 175 h of continuous operation, the vehicle’s unreli-
authors designed the RBD of the vehicle as shown in Fig. 5. ability exceeds its reliability.
Note that HDT is a series system, that is, if any vehi- If 175 h of continuous operation is an interesting time to
cle system fails, the vehicle will do the same. Therefore, analyze the vehicle, the authors implemented the structural

Table 4  Most critical failure System Component Potential failure modes RPN


modes of HDT
Power transmission Cardan couplings Loss of data 448
Electric and control Starter engine Engine does not start or runs slowly 343
Brake and pneumatic Rear and front brakes Brake drum overheating 320
Vibration when brakes are applied 294
Brake drum overheating 288
Power transmission Power takes over Fracture or wear of the shaft teeth 252
Brake and pneumatic Rear and front brakes Vibration when brakes are applied 252
Loss of braking ability 252
Vehicle does not remain in parking position 252
Rear and front brakes Wheel stalls or does not brake 252

Table 5  MTTF and fitted System Distribution Parameter 1 Parameter 2 p-Value MTTF (h)
distributions for the TTF of
each vehicle’s system Electric and control Weibull 0.9645 0.0010 0.0570 1096
Brakes and pneumatic Exponential 0.0009 0.0687 1281
Engine Lognormal 7.2472 0.9072 0.0724 1858
Suspension Weibull 0.8829 0.0023 0.0991 2431
Steering Exponential 0.0003 0.1351 3106
Power transmission Exponential 0.0008 0.0597 3102

The chassis and bodywork system does not present failure data in the horizon time under analysis. In
Weibull distribution, parameter 1 and parameter 2 correspond to shape and scale, respectively. In Lognor-
mal distribution, parameter 1 represents the mean and parameter 2, the standard deviation

Fig. 5  RBD of HDT

13
495 
Page 12 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

Fig. 6  Results of the reliability analysis

IM considering reliability values in this time range to iden- reliability of the HDT, once the RRW importance meas-
tify the most critical systems on the asset. ure establishes the change in unreliability when the unre-
Structural IMs are used when only the structure of the liability of component i is set to zero, that is, by assuming
system is known. Nevertheless, depending on the data and a that a component is ‘perfect’ (or its failure probability is
priori knowledge available about the asset, there are differ- zero).
ent groups of IM that could be analyzed [60]. Table 6 pre- Note that the Birnbaum measure does not consider the
sents the Birnbaum, RRW and RAW importance measures baseline performance (probability of success or failure
obtained for HDT. of an element), making it hard to use it for reliability-
Note that even though every single component is relevant informed decision-making, since low-failure probability
in the HDT operation, the inquiries in the IM field are inter- items are not necessarily the prime candidates for any
esting due to the usual limitation of available resources, and change [38, 60].
the prioritization actions must be assertive to achieve the Therefore, the reliability modeling presented in this
best possible efficiency during the operation. study not only calculates MTBF but also presents the
In this sense, the Birnbaum measure shows that a small probability of failure of a system, according to reliabil-
change in the reliability of electric and control systems ity values, identifying the operational time at which the
results in a large change in the HDT reliability. The RRW cumulative failure probability exceeds the vehicle reli-
measure shows that the engine is the system that should ability and exposes the most critical systems from differ-
be improved to obtain a considerable increase in the ent viewpoints.

Table 6  Structural importance System BI System RRW​ System RAW​


measures of HDT
Electric and control 0.6283 Engine 0.9892 Electric and control 0
Brakes and pneumatic 0.6214 Steering 0.9452 Brakes and pneumatic 0
Power transmission 0.6051 Suspension 0.9024 Engine 0
Suspension 0.5864 Power transmission 0.8746 Suspension 0
Steering 0.5598 Brakes and pneumatic 0.8516 Steering 0
Engine 0.5350 Electric and control 0.8422 Power transmission 0

RAW importance does not offer relevant information about critical HDT systems because all of them are in
series

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 13 of 17  495

Table 7  MTTR and Fitted System Distribution Parameter 1 Parameter 2 p-Value MTTR (h)
distributions for the TTR of
each vehicle’s system Electric and control Lognormal 2.6065 1.9135 0.4511 66
Brakes and pneumatic Lognormal 2.4959 1.7876 0.1270 46
Engine Weibull 0.7236 56.4616 0.9979 71
Suspension Lognormal 2.9191 1.7443 0.4498 65
Steering Weibull 0.6471 43.4425 0.9971 61
Power transmission Lognormal 2.8409 1.7789 0.6884 70

The chassis and bodywork system does not present failure data in the time horizon under analysis. In
Weibull distribution, parameter 1 and parameter 2 correspond to shape and scale, respectively. In Lognor-
mal distribution, parameter 1 represents the mean and parameter 2, the standard deviation

Fig. 7  MCS approach for main-


tainability analysis

Table 8  Repair times of HDT obtained by Monte Carlo simulation


3.3 Maintainability analysis Samples Repair time (h)

As previously said, the inputs required for maintainability Mean Median Mode Percentile 5th Percentile 95th
analysis are the mission time and repair rate of asset com- 1.000 20.44 11.68 2.49 0.66 73.69
ponents/systems. Particularly, the mission time for the HDT 10.000 18.89 10.40 0.59 0.61 67.36
study was defined as 250 operational hours and, to estimate 100.000 18.98 10.58 0.50 0.65 67.17
the repair rate of each vehicle system, a retrospective study 1.000.000 18.60 10.31 1.00 0.65 66.36
was implemented. Table 7 presents the fitted distributions
and MTTR for each system.
Note that, according to Fig. 2, one has various alterna- Carlo analysis using Python and Akram [2] discussed a
tives to model the maintenance process of the asset: MCS, practical method to apply MCS in Excel.
Markov Chain, DFTA, DRBD, DBN, or Petri Nets. How- For the HDT, statistical software R was used to develop
ever, Markov Chains could be complex when the asset under MCS to model repair times of assets. Considering that one
analysis presents different distributions for the transitions usually performs Monte Carlo experiments in long-run
between states (operational and failure) [32, 37, 42]. lengths to increase the results confidence level [65], four
Thus, considering that the asset under study presents dif- different numbers of simulations (samples) were developed.
ferent distributions for both the hazard rate and the repair Table 8 summarizes the simulated repair times of vehicles.
rate of systems, the use of MCS is proposed to model the In general terms, there is a 90% likelihood that the vehicle
maintenance process for the HDT. As shown in Fig. 7, MCS repair time is greater than 0.6 h and less than 66.36 h with
uses the RBD of the asset and the information compiled a repair time average of 18.60 h. Additionally, observing
about the repair and failure of each system to model the situ- Table 8, the results based on 100,000 simulations tend to be
ation in which the HDT is going from the operational state the same as the results from 1,000,000 simulations. Thus,
to the failure state. At each time, at least one of its systems running 1,000,000 simulations to analyze the HDT may not
fails and the vehicle returns to its operational state, each time generate a relevant gain regarding computational cost.
a repair is completed, and all systems are operative. Then, As regards the adequate pdf family to model repair times
MCS allows the maintainability distribution of the asset. of HDT, the generalized additive models for location scale
The MCS method could be implemented in various com- (GAMLSS) available in R [14] was implemented; different
putational tools. Robert and Casella [45] introduced the models that assume a parametric distribution for the interest
MCS in R, Dirk et al. [17] presented the process to imple- variable were applied to obtain the theoretical pdf of HDT
ment MCS in MATLAB, Knox [29] developed a Monte repair times.

13
495 
Page 14 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

Fig. 8  Pdf of HDT repair times

Hence, based on the results from the GAMLSS, it can Table 9  Availability analysis results of HDT
be concluded that the Gamma distribution is an adequate Operational Steady-state MTTR (h) MTBF (h) Inherent
repair time model of HDT as the p-value obtained in the time simulated availability availability
Kolmogorov–Smirnov test is greater than 0.05 for all sam- (h)
ples of MCS. Figure 8 shows the histogram of HDT repair
200 0.9709 27.50 210.57 0.8845
times simulated and the fitted theoretical pdf of each number
400 0.9636 27.23 205.25 0.8828
of simulations.
600 0.9661 27.40 202.58 0.8809
Thus, the maintainability modeling presented herein not
800 0.9681 27.56 191.86 0.8744
only calculates MTTR but also presents the 90th confidence
1000 0.9698 21.63 146.64 0.8715
level for TTR and identifies the pdf family adequate to model
1500 0.9678 32.55 200.11 0.8601
the MTTR system from parametric models.
2000 0.9715 32.40 197.78 0.8593

3.4 Availability analysis
vehicles. Therefore, the times simulated refer to 12, 25, 38,
In the next step in the proposed methodology, the MTBF 50, 63, 94, and 125 operational days.
and MTTR were estimated to quantify the steady-state avail- Considering that the vehicle operates under demand-
ability and inherent availability of the HDT (see Sect. 2). ing conditions, including continuous stops and runs, long
Table  9 presents the availability values obtained from trajectories up to the company’s facilities, steep hills, and
100,000 simulations of 200, 400, 600, 800, 1000, 1500, and frequent load-unload cycles, it is reasonable to conclude
2000 h of operation. Note that the current operating policy that the availability results for the HDT can be considered
of the company establishes 16 daily operational hours for high. Additionally, the inherent availability was lower than

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 15 of 17  495

the steady-state availability for all the operational times was exemplified by means of a case study involving Heavy-
simulated. Duty Trucks. The results showed that the implementation of
This makes sense once the steady-state availability rep- the methodology to analyze the HDT resulted in operational,
resents the asymptotic value of availability, whereas inher- tactical, and strategic planning and support for decision-mak-
ent availability estimation considers that all the downtimes ers. At the operational level, critical failure modes were iden-
occur due to corrective maintenance and the MTBF and tified in the power transmission system and brake and pneu-
MTTR are calculated as: matic systems that can impact the effectiveness of the asset.
Recommendations were generated stating the maintenance
Uptime
MTBF = (4) team should plan a revision of the vehicle before 175 opera-
Number of system failure tional hours since, at that moment, the vehicle’s unreliability
exceeded its reliability. At the tactical level, vehicle operation
Corrective Maintenance downtime planners should expect 96% asymptotic availability with an
MTTR = (5)
Number of system failure MTTR of 18.6 h. At the strategic level, the RAM analysis
showed that the engine should be prioritized during an over-
Even when the inherent availability is higher than 85%,
haul of the asset since by improving this system, the maximum
the HDT availability in the field might be lower than the
increase in vehicle reliability is achieved.
results reported in Table 9 depending on the logistic delays,
In the future, the developed RAM analysis methodology
supply delays and administrative delays.
can include the uncertainty modeling to support the design
strategies to reduce unacceptable risks.

4 Conclusions
References
RAM analysis represents a powerful tool that, if strategically
applied, is capable of increasing asset performance and pro- 1. AIAG & VDA (2012) New AIAG & VDA FMEA handbook,
ductivity, evaluating the impact of failures, and proposing Berlin. Automotive Industry Action Group and Verband der
Automobilindustrie
changes to current operations management. 2. Akram N (2017) Practical Monte Carlo simulation with excel.
The proposed methodology allows acquiring operational Gatekeeper Press, Ohio
and technical knowledge of the asset so that the analyst can 3. Amari SV (2007) Steady-state availability and MTBF of systems
migrate to a more detailed analysis. For instance, the func- subjected to suspended animation. Int J Perform Eng 3:282–284
4. Anderson T, Darling D (1954) A test of goodness of fit. J Am Stat
tional analysis of an asset provides a taxonomic visualization Assoc 49:765–769
that makes it easier to apply a reliability analysis technique 5. Arias Velásquez RM, Mejía Lara JV (2018) Reliability, availabil-
(e.g., RBD, FTA, and BN). ity and maintainability study for failure analysis in series capacitor
Also presented was a detailed flowchart with the method- bank. Eng Fail Anal 86:158–167. https://​doi.​org/​10.​1016/j.​engfa​
ilanal.​2018.​01.​008
ology steps that aim to integrate the results of all techniques 6. Ayele YZ, Barabady J, Droguett EL (2016) Dynamic Bayesian
to allow practitioners to make decisions supported by the network—based risk assessment for arctic offshore drilling waste
different perspectives of an asset. In this sense, analysis tech- handling practices. J Offshore Mech Arct Eng. https://​doi.​org/​10.​
niques such as IM are not only used to provide a ranking to 1115/1.​40337​13
7. Bahri S, Ghribi F, Bacha HB (2009) A study of asymptotic availa-
identify critical components of an asset, but the methodol- bility modeling for a failure and a repair rates following a Weibull
ogy provides support to employ these measures with the distribution. Reliab Risk Anal Theory Appl 2:30–42
functional analysis and FMEA to know how the failure of a 8. Chalabi N, Dahane M, Beldjilali B, Neki A (2016) Optimisation
component might evolve in time. of preventive maintenance grouping strategy for multi-component
series systems: particle swarm based approach. Comput Ind Eng
In terms of computational cost, companies currently have 102:440–451. https://​doi.​org/​10.​1016/j.​cie.​2016.​04.​018
different free alternatives to implement simple codes and 9. Cheung H, Wang S (2019) Reliability and availability assessment
quickly obtain results of a RAM analysis, such as R Studio, and enhancement of water-cooled multi-chiller cooling systems
Python, and MATLAB. In this paper, the software R Studio for data centers. Reliab Eng Syst Saf 191:106573. https://​doi.​org/​
10.​1016/j.​ress.​2019.​106573
was used to implement the RAM analysis due to its statis- 10. Chouaib A (2015) Kernel estimator and bandwidth selection for
tical strengths. Also, there are packages such as FaulTree density and its derivatives. Cornell University, New York, pp 1–22
available to implement basic reliability techniques. 11. Choudhary D, Tripathi M, Shankar R (2019) Reliability, avail-
The authors have used the proposed methodology to analyze ability and maintainability analysis of a cement plant: a case
study. Int J Q Reliab Manag 36:298–313. https://​doi.​org/​10.​1108/​
systems with different levels of complexity, such as Dynamic IJQRM-​10-​2017-​0215
Positioning Systems, used in offshore operations, Offshore 12. Clavijo Mesa MV (2020) Application of reliability, availability
Drilling Rigs Cutting Dryer and Solid Waste Collection and and maintainability analysis to dynamic positioning systems used
Compaction Truck [12, 55]. In particular, the methodology in offshore operations. University of Sao Paulo

13
495 
Page 16 of 17 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495

13. Coit DW, Zio E (2019) The evolution of system reliability optimi- 34. Liberopoulos G, Tsarouhas P (2005) Reliability analysis of an
zation. Reliab Eng Syst Saf 192:106259. https://d​ oi.o​ rg/1​ 0.1​ 016/j.​ automated pizza production line. J Food Eng 69:79–96. https://​
ress.​2018.​09.​008 doi.​org/​10.​1016/j.​jfood​eng.​2004.​07.​014
14. CRAN-R Project (2019) Package GAMLSS [WWW Document]. 35. Marsaglia G, Tsang WW, Wang J (2003) Evaluating Kolmogo-
https://​cran.r-​proje​ct.​org/​web/​packa​ges/​gamlss/​index.​html rov’s distribution. J Stat Softw 8(18):1–4
15. Csorgo S, Faraway J (1996) The exact and asymptotic distributions 36. Martorell S, Sánchez A, Carlos S, Serradell V (2004) Alternatives
of Cramer-von Mises statistics. J R Stat Soc Ser B 58:221–234 and challenges in optimizing industrial safety using genetic algo-
16. Diallo O, Rodrigues JJPC, Sene M (2015) Performances evalua- rithms. Reliab Eng Syst Saf 86:25–38. https://​doi.​org/​10.​1016/j.​
tion and Petri nets, modeling and simulation of computer networks ress.​2003.​12.​010
and systems: methodologies and applications. Elsevier. https://d​ oi.​ 37. Mehta PG, Vaidya U, Banaszuk A (2008) Markov chains, entropy, and
org/​10.​1016/​B978-0-​12-​800887-​4.​00011-0 fundamental limitations in nonlinear stabilization. IEEE Trans Auto-
17. Dirk P, Taimre T, Botev Z (2011) Handbook of Monte Carlo meth- mat Contr 53:784–791. https://​doi.​org/​10.​1109/​TAC.​2008.​917640
ods. Wiley, New Jersey 38. Modarres M (2006) Importance ranking in probabilistic risk
18. Distefano S, Puliafito A (2007) Dynamic reliability block assessment. In: Francis CT (ed) Risk analysis in engineering,
diagrams vs dynamic fault trees. In: Proceedings of the 2007 London, pp 242–247. https://​doi.​org/​10.​1201/​b21429
annual reliability and maintainability symposium, Orlando, FL, 39. Modarres M, Kaminskiy MP, Krivtsov V (2017) Reliability engi-
USA, 22–25 January 2007, pp 71–76 neering and risk analysis: a practical guide, 3rd edn. Taylor &
19. Durga Rao K, Gopika V, Sanyasi Rao VVS, Kushwaha HS, Francis Group, LLC, New York
Verma AK, Srividya A (2009) Dynamic fault tree analysis using 40. Montgomery DC, Runger GC (2003) Applied statistics and prob-
Monte Carlo simulation in probabilistic safety assessment. ability for engineers. European Journal of Engineering Education.
Reliab Eng Syst Saf 94:872–883. https://​doi.​org/​10.​1016/j.​ress.​ New York. https://​doi.​org/​10.​1080/​03043​79940​89283​33
2008.​09.​007 41. NASA (2002) Fault tree handbook with aerospace applications.
20. Ebeling, C., 1997. An introduction to Reliability and Maintain- NASA Handb XXXIII:218. https://​doi.​org/​10.​1109/​TC.​1986.​
ability Engineering. Massachusetts. 16768​19
21. Gao S, Wang J (2020) Reliability and availability analysis of 42. Osorio-Lird A, Chamorro A, Videla C, Tighe S, Torres-Machi C
a retrial system with mixed standbys and an unreliable repair (2018) Application of Markov chains and Monte Carlo simula-
facility. Reliab Eng Syst Saf 205:107240. https://​doi.​org/​10.​ tions for developing pavement performance models for urban net-
1016/j.​ress.​2020.​107240 work management. Struct Infrastruct Eng 14:1169–1181. https://​
22. Garg H (2014) Reliability, availability and maintainability doi.​org/​10.​1080/​15732​479.​2017.​14020​64
analysis of industrial systems using PSO and fuzzy methodol- 43. Possan E, De Oliveira Andrade JJ (2014) Markov chains and reli-
ogy. Mapan—J Metrol Soc India 29:115–129. https://​doi.​org/​ ability analysis for reinforced concrete structure service life. Mater
10.​1007/​s12647-​013-​0081-x Res 17:593–602
23. Giakoumis E (2016) Driving and engine cycles. Springer, New 44. R Core Team (2016) R: a language and environment for statisti-
York cal computing. R Foundation for Statistical Computing, Vienna,
24. Guevara Carazas FJ, Martha De Souza GF (2009) Availability Austria. https://​www.R-​proje​ct.​org/
analysis of gas turbines used in power plants. Int J Thermodyn 45. Robert C, Casella G (2010) Introducing Monte Carlo methods
12:28–37. https://​doi.​org/​10.​5541/​ijot.​10340​00235 with R, New York
25. Guo H, Yang X (2007) A simple reliability block diagram 46. Robidoux R, Xu H, Xing L, Zhou M (2010) Automated modeling
method for safety integrity verification. Reliab Eng Syst Saf of dynamic reliability block diagrams using colored Petri nets.
92:1267–1273 IEEE Trans Syst Man Cybern Part A Syst Hum 40:337–351
26. Herder PM, van Luijk JA, Bruijnooge J (2008) Industrial appli- 47. SAE International (2001) Recommended failure modes and effects
cation of RAM modeling. Development and implementation of analysis (FMEA) practices for non-automobile applications. Soci-
a RAM simulation model for the ­Lexan® plant at GE Industrial. ety of Automotive Engineers International, New York
Plast Reliab Eng Syst Saf 93:501–508. https://​doi.​org/​10.​1016/j.​ 48. Saini M, Kumar A (2019) Performance analysis of evaporation
ress.​2006.​10.​019 system in sugar industry using RAMD analysis. J Braz Soc Mech
27. ISO (2016) Petroleum, petrochemical and natural gas industries— Sci Eng 41:1–10. https://​doi.​org/​10.​1007/​s40430-​019-​1681-3
collection and exchange of reliability and maintenance data for 49. Seo J, Kim H, Park S (2018) Estimation of CO2 emissions from
equipment (ISO 14224:2016) heavy-duty vehicles in Korea and potential for reduction based on
28. Kabir S (2017) An overview of fault tree analysis and its appli- scenario analysis. Sci Total Environ 636:1192–1201. https://​doi.​
cation in model based dependability analysis. Expert Syst Appl org/​10.​1016/j.​scito​tenv.​2018.​04.​269
77:114–135. https://​doi.​org/​10.​1016/j.​eswa.​2017.​01.​058 50. Sharma A, Yadava GS, Deshmukh SG (2011) A literature review
29. Knox A (2011) Multivariable Monte Carlo analysis methods and future perspectives on maintenance optimization. J Q Maint
in traffic accident reconstruction using python. In ASME 2011 Eng 17:5–25. https://​doi.​org/​10.​1108/​13552​51111​11162​22
international mechanical engineering congress and exposition, 51. Sharma RK, Kumar S (2008) Performance modeling in critical
Colorado, pp 639–653 engineering systems using RAM analysis. Reliab Eng Syst Saf
30. Kumar P, Jain M (2020) Reliability analysis of a multi-component 93:913–919. https://​doi.​org/​10.​1016/j.​ress.​2007.​03.​039
machining system with service interruption, imperfect coverage, 52. Si XS, Wang W, Hu CH, Zhou DH (2011) Remaining useful life
and reboot. Reliab Eng Syst Saf 202:106991. https://​doi.​org/​10.​ estimation—a review on the statistical data driven approaches.
1016/j.​ress.​2020.​106991 Eur J Oper Res 213:1–14. https://​doi.​org/​10.​1016/j.​ejor.​2010.​11.​
31. Lewis E (1996) Introduction to reliability engineering. Wiley, 018
New York 53. Sikos L, Klemeš J (2010) Reliability, availability and mainte-
32. Li W, Zhang C (2009) Markov chain analysis. Int Encycl Hum nance optimisation of heat exchanger networks. Appl Therm Eng
Geogr. https://​doi.​org/​10.​1016/​B978-​00804​4910-4.​00470-3 30:63–69. https://​doi.​org/​10.​1016/j.​applt​herma​leng.​2009.​02.​013
33. Li H, Guedes-Soares C, Huang HZ (2020) Reliability analysis of 54. Silverman B (1986) Density estimation for statistics and data
a floating offshore wind turbine using Bayesian Networks. Ocean analysis. Monogr Stat Appl Probab 33:43–54. https://​doi.​org/​10.​
Eng 217:107827 1177/​03009​85896​03300​105

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2021) 43:495 Page 17 of 17  495

55. Souza-Franco V, Clavijo MV, Schleder A, Martins MR (2019) 61. Van Eck NJ, Waltman L (2009) VOSviewer: a computer program
Monte Carlo simulation to consider uncertainty in the reliability for bibliometric mapping. In: 2th international conference on sci-
analysis of dynamic positioning systems. In: Proceedings of the entometrics and informetrics, pp 886–897
29th European safety and reliability conference, 22–26 September 62. Xing L, Amari SV (2008) Fault tree analysis. In: Handbook
2019, Hannover of performability engineering. Springer, London, pp 595–620.
56. Stapelberg R (2009) Handbook of reliability, availability, main- https://​doi.​org/​10.​1007/​978-1-​84800-​131-2_​38
tainability and safety in engineering design. Springer, London. 63. Yang DY, Tsao CL (2019) Reliability and availability analysis
https://​doi.​org/​10.​1007/​978-1-​84800-​175-6 of standby systems with working vacations and retrial of failed
57. Straub D, Papaioannou I (2015) Bayesian updating with structural components. Reliab Eng Syst Saf 182:46–55. https://​doi.​org/​10.​
reliability methods. J Eng Mech. https://​doi.​org/​10.​1061/(​ ASCE)​ 1016/j.​ress.​2018.​09.​020
EM.​1943-​7889.​00008​39 64. Zhang J, Haskins C, Liu Y, Lundteigen MA (2018) A systems
58. Tsarouhas P (2018) Reliability, availability and maintainability engineering–based approach for framing reliability, availability,
(RAM) analysis for wine packaging production line. Int J Qual and maintainability: A case study for subsea design. Syst Eng
Reliab Manag 35:821–842 21:576–592. https://​doi.​org/​10.​1002/​sys.​21462
59. U.S. Department of Health and Human Services (2006) Guidance 65. Zio E (2013) The Monte Carlo simulation method for system
for industry, quality risk management. Center for Drug Evalua- reliability and risk analysis. Springer, London. https://​doi.​org/​10.​
tion and Research (CDER), Center for Biologics Evaluation and 1007/​978-1-​4471-​4588-2
Research (CBER), Maryland
60. Van Der Borst M (2001) An overview of PSA importance meas- Publisher's Note Springer Nature remains neutral with regard to
ures. Reliab Eng Syst Saf 72:241–245. https://​doi.​org/​10.​1016/​ jurisdictional claims in published maps and institutional affiliations.
S0951-​8320(01)​00007-2

13

You might also like