Elias
Tadeu da Silva
University
of Araraquara, Brazil
E-mail: eng.eliastadeu@bol.com.br
Jorge
Alberto Achcar
University
of São Paulo, Brazil
E-mail: achcar@fmrp.usp.br
Claudio
Luis Piratelli
University
of Araraquara, Brazil
E-mail: clpiratelli@uniara.edu.br
Submission: 10/17/2019
Revision: 1/7/2020
Accept: 1/18/2020
ABSTRACT
Maintenance and management have substantial importance in the search of the company’s competitive advantages. In this direction, a reliability analysis of equipments is very important for the definition of the most suitable maintenance strategy. The main goal of this paper is to assess the reliability-centered maintenance of the industrial reliability curve of a sugar cane department of an industry located in São Paulo State, Brazil. The proposed research method was based on the application of existing statistical modeling for the times between failures (TBF) of all reported equipment failures or the interruption times in the production line. These times were modeled by standard lifetime probability distributions as log-normal and Weibull distributions. The results showed that the company's strategy for preventive maintenance during the off-season is not adequate and the statistical analysis also identified important factors that affect the company's maintenance strategy. These results could be of great interest for the company and for engineering applications in general.
Keywords: maintenance, management, reliability, lifetime, sugar-cane
1.
INTRODUCTION
Competition between
process industrial organizations requires proper maintenance planning to lead
to the most high reliability for the industrial equipment’s, which
significantly could reduce losses during production (RUSCHEL, SANTOS, LOURES, 2017).
For Gandhare,
Akarte and Patil (2018), a
process industry requires large amounts of investment and continued
availability of its facilities, which makes maintenance a critical area of the
company, as it is responsible for the equipment performance. The sugar industry
considered in this study operates seasonally and the availability of its
facility during the sugar-manufacturing period determines its performance
(GANDHARE; AKARTE; PATIL, 2018)
For Alsyouf
(2009), maintenance activities are becoming more complex, as a conventional
manufacturing system does not only consist of traditional mechanical equipment,
but it also incorporates in its system, electronic, hydraulic,
electromechanical, software and human elements.
Inadequately maintained or neglected industrial installations will eventually
require costly repairs, as overtime equipment, machinery or installations will
wear out (VISHNU; REGIKUMA, 2016).
Also according to
Vishnu and Regikuma (2016), the main goal of
maintenance in an industrial plant is to achieve minimum downtime and to keep
the equipment in operating condition at the lowest possible cost. The European
Federation of National Maintenance Societies (EFNMS) defines maintenance as the
combination of all technical, administrative and managerial actions used during
the life cycle of equipment that are intended to retain or restore it to the
state it can satisfactorily perform its required function. Mobley (2002)
points out that maintenance could be classified as: corrective maintenance,
preventive maintenance, predictive maintenance, total productive maintenance
(TPM) and reliability centered maintenance (RCM).
According to Farrerro, Tarrés and Losilla (2002),
in order to obtain an efficient maintenance policy, a combination of
corrective, preventive and predictive maintenance is necessary, but the type of
maintenance utilized and the interval between them is a function of their
behavior, failure rate and the overall
cost involved in the occurred damage.
Fogliatto and Ribeiro
(2008) define RCM (Reliability Centered Maintenance) as a program that combines
maintenance engineering techniques with systematic treatment and its objective
is to guarantee the original function of the manufacturing equipment (see also,
MOUBRAY,1997).
It is important to point out that the sugar and alcohol
production industries have as their main priority the planned corrective
maintenance due to the seasonal characteristics of their operation, since there
is great availability of time in the period known as off-season (approximately
six months). In the food industry, the production line consists of machines
interconnected by common transfer mechanisms with different failure modes. In the event of random failure in some
equipment, most of the line interrupts the process, where the unfinished
product should be discarded because of deterioration or quality problems, that
is, the impact of the failure is negative and causes a decrease in reliability
and production (TSAROUHAS, 2012).
This study is applied in nature, with a
quantitative approach and, as its method, it uses statistical modeling to
estimate the reliability through time-to-failure modeling - TBF of the
production equipment. Equipment operating out of optimum condition can lead to
unrecoverable losses for companies in competitive markets. In general,
investment in production equipment is high, which is why one should seek to
maximize its utilization and hence increase its financial return for the
organization (MENGUE; SELLITTO, 2013; RAPOSO, 2011).
The paper is organized as follows: in sub-section
1.1, the problem and the research questions are introduced; in sub-section 1.2,
the research goals are presented; in sub-section 1.3, the methodological
aspects are delineated; in section 2, the company characterization is
introduced; in section 3, some concepts on reliability models, especially the
Weibull and the log-normal distributions are presented; in subsection 3.4, RCM
applications for maintenance strategies are discussed; in section 4, the
results of the study are presented; finally, in section 5, some discussion on
the obtained results, future work and conclusions are presented.
Sellitto (2005), considering a qualitative study with
several managers, concluded that maintenance strategies are usually elaborated
by subjective methods, that is, in critical equipment they are submitted to
preventive maintenance; for inactive equipment it is allowed emergency
maintenance and for redundant equipment it is recommended corrective
maintenance. For Alsyouf (2007), an effective
maintenance policy makes a company more competitive in the market, through
better use of equipment time.
Maintenance
is usually under increasing pressure to improve the plant's industrial
performance and short-term cost savings, but in the long term, the effects of improper maintenance are
negative and those responsible for maintenance must be able to convince senior
management that a solid investment and structured maintenance will save the
company considerable amounts in the future (WAEYENBERGH; PINTELON, 2009).
RCM is one of
the widely known maintenance strategies used to preserve the operational
efficiency of industrial plants in a variety of sectors, including critical
sectors such as power, military, aviation, rail, oil and naval (CARRETERO et
al. 2003).
In a study by
Vishnu and Regikumar (2016), in the sugarcane
processing industry in India, it was found that RCM in general had not been
implemented due to the lack of adequate methodologies and tools. The policy of
maintaining a sugar cane processing plant, like in this study, is historically
divided into two periods known as harvest and off-season. During the harvest
period, from the beginning of April to the end of November, the plant operates
fully and steadily, with 6-hour periodic monthly shutdowns reconciled with a
planned daily maintenance routine. In this period, also there is the planning
of the maintenance for the off-season period, since the data and demands
collected in this phase are the basis for future maintenance.
In the
off-season period, ranging from the beginning of December to the end of March,
the operation of the industry is totally stopped, due to the end of the
sugarcane harvest. Therefore, the industry is available in this period for
maintenance according to the planning and budget previously established during
the harvest period.
Mengue
and Sellito (2013) define the maintenance strategy
(preventive, predictive, corrective or emergency) most suitable for a
centrifugal pump of an oil plant based on the concepts of reliability theory.
The times obtained in the study were modeled by standard existing probability
distributions and from the obtained inference results it was estimated the
Reliability R(t), Maintainability M(t) and Pump Availability functions (MENGUE;
SELLITO, 2013).
In a related study, Komninakis (2017) evaluated the coherence of the
maintenance strategy of a food industry through statistical modeling applied to
the repair times (TTR) and to the times between failures (TBF) of a production
line consisting of six packaging machines modeled by Log-normal and Weibull
probability distributions. Based on what
it was exposed above by many authors, the main goal of this study is to
evaluate through the proposed studies of Sellitto
(2005), Mengue and Sellitto
(2013) and Komninakis (2017) whether the current
strategy of the maintenance of the company object of study is best suited for
the maintenance management of the sugar cane plant.
The main goal
of this study is to evaluate through Reliability Centered Maintenance
(RCM) whether the current strategy of
the equipment maintenance of a sugar, ethanol and electric power plant is
adequate. Besides the main goal, the study also has as some specific
goals, the analysis of the repair times
(TTR), to estimate the availability of the equipment’s, to position the equipment’s
in the life cycle curve (bathtub curve), to discover specific factors that can
significantly affect the repair times and the times between failures and,
according to the obtained results to suggest to the company managers the best
maintenance strategy to be used in future.
The research used in this paper is descriptive
because it aims to develop a statistical model that describes the faults that
occur in the equipment of an industry and to develop a profile with its
characteristics. This analysis is performed through the information contained
and stored in a company’s own database. Data collection is obtained by
extracting historical data of change of rotation and grinding stops available
in the management system and the database of the company under study. The approach
is quantitative and the method used will be the statistical modeling with the
estimation of the equipment reliability by modeling the times between failures
(TBF), repair times (TTR) and the survival times of the industrial equipment of
a sugar and ethanol producing industry (RAUSAND, 1998; LAFRAIA, 2001).
2.
COMPANY CHARACTERIZATION
The studied company operates in the
sugar and ethanol
industry sector with the production of sugar, ethanol and electricity
generation. The plant was founded in 1953 in the countryside of São Paulo
State, Brazil, together with a colony that houses employees working in the
agricultural and industrial sectors. In the 1990’s, the company was
incorporated by a large group of mills and now it is currently part of a joint
venture between a domestic and a foreign company.
Over the years, it has undergone
numerous renovations and extensions. Its industrial park has a daily sugarcane
crushing capacity of 7,300 tons, producing 20,000 bags of sugar and generating
4.4 MW of electricity. The industry has about 120 employees, 30 of which who
are dedicated exclusively to the maintenance of the industrial plant that operates
24 hours a day, 7 days a week. The industry, as it is the tradition and history
of the sector, operates seasonally two
periods of the year known as harvest and off-season periods.
The maintenance policy of the
company under study is divided into these two periods. The first one is the
off-season period between the months of December and the end of March, during
which the industry's operation is completely paralyzed and available for
maintenance according to the previously established planning and budget.
The off-season period is
particularly warm with rainy climate conducive to sugarcane germination and
growth, and at this time the harvesting of sugarcane is not recommended, as the
excess of soil moisture facilitates the removal of roots during cutting and
makes heavy machinery traffic difficult, damaging sugarcane fields and
compromising future sprouts.
The second period known as the crop
period, occurs in the remainder of the year (April to November), when the plant
operates fully and steadily, similar to a consumer goods company. The company
has as maintenance policy a monthly shutdown either lasting six hours or when
rain occurs, which prevents sugarcane cutting as it was already mentioned; in
the latter case, the duration is indefinite and may reach days depending on the
amount of rain. In this case, the return of industrial activity occurs only
after the resumption of agricultural activity of cutting, loading and
transporting sugarcane to the industry.
The data collected for the
statistical analysis consists of 1209 company fault records from May 1, 2012 to
October 15, 2017. These records contain the times between failures and repair
times related to different equipment.
3.
USE OF RELIABILITY MODELS
The reliability function is one of
the main probabilistic functions used to describe survival studies and is
defined as the probability of an observation not failing until a certain time
t, that is, it is possible to determine
the probability of non-failure successes over a given time (see, for example, GIOLO;
COLOSIMO, 2006). That is, the reliability function is defined by R(t) = P(T > t), t > 0. As a
consequence, the cumulative failure distribution function is defined by F(t) =
P(T ≤ t) = 1 - R(t). The failure probability density function denoted by
f(t) allows the probability of failures to be determined over a period of time
(see, for example, LAWLESS, 1982; ELSAYED, 1995).
Assuming that the reliability
function is derivable and continuous with respect to the failure times, the
cumulative failure distribution function is also derivable. Under this
hypothesis, the cumulative failure distribution function F(t) can be derived to
obtain the failure probability density function f(t). The risk rate or failure
rate function denoted by h(t) is the probability of failure occurring within a
time interval [t1, t2], since it did not occur until t1,
in other words, representing the proportions of failures occurring per unit of
time. The probability of failures in the interval [t1, t2]
can be expressed in terms of the reliability function as R(t1) - R(t2).
Thus, the failure rate in the interval [t1, t2] is
expressed by,
[R(t1) - R(t2)]/(t2-t1)R(t1) [1]
In general, one can represent the interval [t1, t2]
by (t, t + Dt), that is, t2 = t + Dt; thus the rate function is given by h(t) = [R(t) – R (t + Δt)]/[ΔtR(t)]. Assuming
a very small value for Δt , h (t) represents the
instantaneous failure rate, or risk rate, at time t conditional on survival up
to time t, that is, it describes how the instantaneous failure rate changes
over time (see, for example, LAWLESS, 1982).
Thus, one can find from expression (1), when Dt → 0, a very useful formula for the
risk function h (t) given by h(t) = f(t)/R(t). The mean time denoted by MTTF
(mean time to failure) measures the time of an item, component, or system
surviving before failure, that is, the average lifetime MTTF is obtained by the
area under the reliability function, that is,
[2]
3.1.
Reliability Models
In survival analysis or reliability
analysis, there are two types of models: non-parametric models and parametric
models. The use of non-parametric methods allows us to gain perspective on the
nature of the data distribution from which it was designed without, however,
selecting a specific probability distribution (LEWIS, 1994). For Giolo and Colosimo (2006), the
use of parametric techniques has been more frequent in the industrial area than
in the medical area. Although there is a wide variety of probabilistic models
used in the survival analysis, some models gain a prominent position, as they
have proven adequacy in different situations. In this case, we have the exponential
model, Weibull and the lognormal models (GIOLO; COLOSIMO, 2006).
3.2.
Weibull Distribution
The Weibull distribution, widely used in
reliability because of its flexibility in accommodating different forms of risk
function, is perhaps the most widely used distribution model for lifetime
analysis. For a random variable T with Weibull distribution, the probability
density function is given by,
f(t) = exp{- ()γ}, [3]
for
t > 0 where t is the time to failure, γ is
the shape parameter and θ is the scale parameter, all positive. For the
Weibull distribution (3), the survival or reliability function R(t) is given by
R(t) = exp{-(t/θ)γ} and has a
failure rate (hazard function) given by,
h(t)
=
[4]
for t > 0, γ >
0 e θ > 0.
The shape of the survival curve is related to the
parameter γ. With γ > 1, the failure rate is increasing; with
γ = 1, there is a constant failure rate (exponential distribution); with
γ <1, the failure rate is decreasing. The mean life time E(T) and
variance Var(T) of the Weibull model are given
respectively, by,
E(T) =
θ
Γ[1 + ()]
and
Var(T) = θ2{Γ[1 + ()]- Γ[1 + ()]2}, [5]
where
the gamma function, Γ(k), is defined by Γ(k) = exp{-x}dx.
Sellitto (2005)
relates the phases of the bathtub curve life cycle to the values
of the Weibull shape parameter γ, which represents the
behavior of the equipment fault curve, listing the most common types of faults
found in each phase, namely:
· In
the infant mortality phase, where γ < 1, the failure rate is high but
decreasing over time; thus, failures are premature, usually caused by
deficiencies in the manufacturing process, improper installation, or
out-of-specification materials;
· In
the maturity phase, where γ = 1, the failure rate fluctuates around a
constant average, the failures are haphazard and due to less controllable
factors such as equipment misuse, resistance overrun or unpredictable natural
phenomena.
· In the senile mortality or wear phase, where
γ > 1, the failure rate is increasing. Thus, failures are caused by
aging, mechanical, electrical or chemical degradation, fatigue, corrosion, or
very short design life. It is the end of equipment life.
A random
variable T defined for positive values has a lognormal distribution if the
logarithm of T , that is, ln(T) is normally distributed with mean and standard
deviation given respectively by μ and σ2. The probability
density function for T is given by:
[6]
The mean and variance of T are given
respectively by,
[7]
The choice of
a company's maintenance strategy should be driven by the use of quantitative
methods where in recent years several approaches of this type have been studied
and applied to maintenance in companies of various segments such as the
Reliability-Centered Maintenance (RCM) identifying the risks and impacts of
failure modes and thus proposing the best type of maintenance to be performed
to minimize damage (SELLITTO, 2007).
To formulate
an industrial maintenance policy in the metalworking sector, Sellitto (2005) reviewed the concepts related to random
process variables as a way to define the basis of reliability and modeling
applied to maintenance management through the modeling of the techniques from
time to failure and until repair was established the maintenance policies in
factories of this sector (SELLITO, 2005).
Mengue
and Sellito (2013) defined the maintenance strategy
(preventive, predictive, corrective or emergency) most suitable for a centrifugal
pump of an oil plant based on the concepts of reliability. The times obtained
in the study were modeled by probability distributions and, from the obtained
results, the estimated reliability functions R(t), Maintainability M(t) and
Pump Availability (MENGUE; SELLITO, 2013) were estimated. Komninakis
(2017) evaluated the coherence of the maintenance strategy of a food industry
through statistical modeling applied to the repair time (TTR) and time between
failures (TBF) of a production line consisting of six packaging machines
modeled by Log-normal and Weibull probability distributions (see, KOMNINAKIS; PIRATELLI; ACHCAR, 2018).
Chopra, Sachdeva and Bhardwaj (2016)
studied the relationship between MCC implementation factors and increased
productivity in the Indian process industries of the textile, fertilizer,
pharmaceutical, food and beverage industries. For this purpose, a questionnaire
was prepared containing questions to be answered on the four-point Likert scale
(1- nominal gain, 2- reasonable gain, 3- high gain and extremely high 4-gain)
and delivered to 100 small, medium and large companies with 64 responding
companies. The questionnaire was divided into two sections: the first dealt
with RCM implementation factors and the second dealt with manufacturing
parameters, that is, productivity. The obtained result showed that companies
with higher level of MCC implementation have higher productivity compared to
companies with lower level of implementation.
Gandhare, Akarte
and Patil (2018) conducted an empirical
investigation of maintenance performance management practices in the sugar
processing industry in India. Through the collected data, statistical methods
such as correlation, multiple regression and cluster analysis were used to
achieve the objective of the study which was to understand the used maintenance
practices and the differences in maintenance performance among the industries
analyzed in the survey.
To verify the
best decision-making process for maintenance on equipment in power plants with
gas turbines (GTPP), steam turbines (STPP) and combined cycle (CCPP), Sabouhi Abbaspour, Fotuhi-Firuzabad and Dehghania (2016)
reported the use of reliability theory to quantify the criticality and
importance of each individual component in the reliability performance of a
power generation system. For this purpose it was necessary to identify the
series and parallel arrangements of the components and later to elaborate an
analysis of the overall system reliability indexes such as the repair rate,
average system cycle time (SCT), average system downtime (TMI), mean time to premature failure (MTTF1) and
mean time to failure (MTTF). This approach helped operators and managers
understand the importance of each component in the overall plant performance
and to show that the STPP system is more reliable, followed by the GTPP and
CCPP ones.
Vishnu and Regikuma (2016) developed a general RCM model that is
suitable for all types of process plants with interconnected complex subsystems
and critical components. For this purpose, a framework was developed following
the methodology based on the Analytical Hierarchy Process (AHP) to develop a
database system that monitors maintenance actions and equipment information to
define cost-effective ways to increase industry availability and profitability.
Heo, Kim and Lyu (2014) presented a Reliability Centered
Maintenance (RCM) model to analyze the power sector maintenance strategy to
maintain reliability also studying potential failures in substation
transmission system components. The goal of the study was to find an optimal
maintenance strategy and to compare MCC with the current adopted strategies,
that is, the Time Based Maintenance and the Condition Based Maintenance. The
obtained results showed that MCC has a best cost-benefit ratio than the others.
4.
RESULTS
The data collected for the statistical
analysis consist of 1209 company fault records from May 1, 2012 to October 15,
2017, divided into 6 crop years. These records contain the times between
failures (TBF) and repair times (TTR) related to different equipment that may
interfere with the industry shutdown, that is, at the time the milling process
is interrupted and the milling is stopped. In this study, only the results of
the statistical analysis for the TBF data denoted here by MTBF are presented.
Initially, an ANOVA
(analysis of variance model) model is considered in the data analysis to
compare the TBF means in different years. In this case, the data are considered
in the logarithmic scale since the transformed data presents better normality
compared to the data in the original scale. This is observed in the normal
probability plots given in Figure 1 for the 1209 observations considered in the
study; similarly, good normality is also observed if we analyze each year separately, that is,
the normality is better when using log-scale data.
|
|
Figure 1: Normality
plots for the original and logarithmic scale data.
Source: The authors
To verify
statistically whether there are significant differences among the MTBF means
for the different years, this section considers the use of a one-way analysis
of variance model. The analysis of variance technique (ANOVA) is a statistical
methodology to test if a given factor has a significant effect on the dependent
variable denoted by Y (MONTGOMERY; RUNGER, 2010).
The
ANOVA results with one classification were obtained using Minitab®
software to compare the MTBF means on the logarithmic scale. Since the obtained
p-value is smaller than 0.001, that is, much lower than 0.05 (the usual
significance level), there is significative
statistical difference among the means of MTBF in different years.
The
needed assumptions to validate the inferences in the ANOVA model were verified
from residual graphs (normality and constant variance of residuals). From the
plots of Figure 2 (95% confidence intervals for the means), significant
differences among the means for the different years are also confirmed (95%
confidence intervals for the means are not overlapping).
In this section, the TBF data are analyzed
by reliability models assuming the original scale data to verify the
maintenance performance between years. From Weibull and log-normal probability
graphs obtained from the models fitted by the maximum likelihood estimator
(MLE) methods and the Minitab® software, it is observed that the
Weibull distribution is better fitted by
the data (points closest to the line in the Weibull probability graph when
compared to the normal probability graph) as observed in Figure 3.
Thus, the Weibull
distribution is assumed in order to obtain the reliability curves estimated by
the maximum likelihood method for the MTBF times in the different years
reported in the database. The failure rates, reliability functions and MLE
estimators of the shape and scale parameters are presented in Figure 4.
Figure 2: 95%
confidence intervals for log averages (MTBF) in different years.
Source:The authors (2019)
Figure 3:
Weibull and Lognormal probability graphs.
Source: The authors (2019)
In Figure 5 (extracted values of Figure 4) the
shape parameter estimators (γ) for
the Weibull fit are presented for the MTBF data in each year. From these values, it is observed
that for all cases there are always values γ < 1, which does not
present a pattern that can relate to a possible improvement, that is, with a
trend γ ≅ 1, phase of maturity in the life cycle.
Figure 4:
Failure rate and reliability graphs assuming a Weibull distribution
Source: The authors (2019)
Figure 5:
Weibull shape parameter estimator versus year
Source: The authors (2019)
In this section, it is compared the MTBF
means for the different months, using a one-way analysis of variance model
using the Minitab® software with the data in the logarithmic scale.
Since the obtained p-value is smaller than 0.001, there is statistical
difference among the means of MTBF in different months.
Figure 6: 95%
confidence intervals for the mean of MTBF in each month
Source: The authors (2019)
Normality
and constant variance of residuals also were verified by standard residual
plots. From the plots of Figure 6 (95% confidence intervals for the means),
significant differences among the means for the different months also are
confirmed (95% confidence intervals for means are not overlapping).
From
the obtained results, it is observed that the mean of MTBF is larger for
October when compared to the other months. In the same way, a smaller estimate
for the mean of MTBF for December is observed.
In this
section the TBF data are analyzed by reliability models assuming the original
scale data in order to verify the maintenance performance among different
months. From Weibull
and log-normal probability graphs obtained from the models fitted by the
maximum likelihood estimator (MLE) methods and the Minitab®
software, it is observed that the Lognormal distribution is better fitted by the data (points closest to the line in
the Normal probability graph when compared to the Weibull probability graph) as
observed in Figure 7.
Thus, the Lognormal distribution is
assumed to obtain the reliability curves estimated by the maximum likelihood
method for the MTBF times in the different months reported in the database. The
failure rates, reliability functions and MLE estimators of the shape and scale
parameters are presented in Figure 8.
Figure 7:
Weibull and Lognormal probability graphs.
Source: The authors (2019)
Figure 8:
Failure rate and reliability graphs assuming a Lognormal distribution
Source: The authors (2019)
To
verify statistically if there is significant differences between the MTBF means
for the different causes of failure, this section considers the use of a
one-way analysis of variance model using Minitab® software to
compare the MTBF means on the logarithmic scale. Since the obtained p-value is
smaller than 0.001, there is statistical difference among the means of MTBF in
different causes of failure. Normality and constant variance of residuals also
were verified by standard residual plots. From the plots of Figure 9 and Table
1 (95% confidence intervals for the means), significant differences between the
means for the different causes are also confirmed (95% confidence intervals for
means are not overlapping).
From the obtained results, it is
observed that the mean of MTBF is larger for civil maintenance (7.534),
although there is only two observations, when compared to the other causes.
Table 1:
Estimated means for MTBF and 95% confidence intervals for the mean of MTBF in each different cause
Cause
|
N
|
Mean
|
95% CI |
Excess Capacity Milling |
1 |
5.704 |
(2.261; 9.146) |
Maintenance Industrial Automation |
9 |
6.242 |
(5.095; 7.390) |
Civil Maintenance |
2 |
7.534 |
(5.100; 9.969) |
Maintenance Instrumentation |
15 |
6.178 |
(5.289; 7.067) |
Mechanical Maintenance |
290 |
6.0876 |
(5.8855; 6.2898) |
Operating - Sugar Factory |
258 |
6.2799 |
(6.0656; 6.4942) |
Operating - Power Generation |
5 |
6.429 |
(4.890; 7.969) |
Operating - Steam Generation |
135 |
4.924 |
(4.628; 5.220) |
Operating - Preparation / Grinding |
276 |
6.477 |
(6.269; 6.684) |
Electrical Maintenance |
218 |
5.564 |
(5.330; 5.797) |
Source: The authors (2019)
5.
GENERAL CONCLUSIONS AND FUTURE WORK
The main goal of the study was to verify if the maintenance strategy adopted by
the company is the most appropriate to manage its maintenance. Related to four
different factors, we have some conclusions and interpretations:
· Regarding to the year factor, it was
possible to identify that, in the bathtub curve, the sector is in the infant
mortality phase (γ <1) and, therefore, should use the most appropriate
maintenance strategy for this phase. Table 2 shows a comparison between the
recommended maintenance practices for infant mortality according to Sellitto (2005) and that practiced by the company. In the
region of infant mortality, premature failures occur due to errors in
manufacturing processes, installation or application of equipment materials
(MENGUE; SELLITTO, 2013). According to Table 2, the recommended maintenance
strategies were corrective and emergency, as they would seek the root cause of
possible defects and eliminate them from operation. However, the sugar and
alcohol industry sector, as well as the analyzed company, has the practice of
supporting its decisions in the employees' experiences and in the history of
breakages and, thus, makes intensive use of preventive maintenance during the
off-season, through the massive replacement of static and rotating items
(electrical and mechanical) without a well-defined criterion by inspections and
technical reports. Thus, the company was mistaken in the strategy, because, as
mentioned in Table 2, the practice of preventive maintenance perpetuates and
even aggravates mortality, as it replaces the survivors of the previous crop,
instead of preventing the breakdown as expected.
Table 2: Life
cycle maintenance strategy of the infant mortality and company mortality
lifecycle practiced by the company (infant mortality, failures in origin)
Estrategy |
Consequence |
Recommended |
Company |
Emergency |
Delays or
even prevents the end of child mortality by not reinforcing items that have
broken or not removing causes of origin failures |
Yes |
Yes |
Corrective |
Anticipates
the end of child mortality by reinforcing items that have broken or removed
the causes of origin failures |
Yes |
Yes |
Preditive |
Monitors
failures in progress that can result in breakage, but these are very few at
this stage, as breaks give more for low resistance |
No |
No |
Preventive |
Perpetuates
or even aggravates child mortality by accurately exchanging survivors, strong
items that have no source flaws |
No |
Yes |
Source: The
authors (2019)
· Still regarding the year factor, the
highest MTBF average was observed for the year 2016. It was also observed that
the MTBF average increases during the period from 2012 to 2016, but in 2017 there is a decrease in the MTBF average. This drop is due to the continuity
of the wrong maintenance practices, which initially showed favorable results,
but were not sustainable for long periods.
· Regarding the
month factor, from the obtained results, it was observed that the MTBF average
is higher for October, when compared to the other months. Similarly, a lower
estimate for the MTBF average is observed for December. This is explained when
the data are analyzed chronologically, between the beginning (April and May)
and end of harvest (November and December) which are more subject to
disturbances that lead to the interruption of the industrial operation. The
start months are susceptible to process variations and adjustments and the end
months to sharp wear and tear of equipment throughout the crop year resulting
in equipment breakdown. The remaining months (mid-season) have historically
been stable due to few variations and continuous operation of the industrial
plant.
·
Regarding The cause of failure factor,
from the obtained results, it was observed that the MTBF average is higher for
civil maintenance (7.534) when compared to the other causes. This fact is the
result of the total demobilization of the civil maintenance sector of the
industrial plant, through the strategic definition of the company. Thus, in the
occurrence of any event, the emergency hiring of external labor is necessary,
which makes any type of repair of this order too slow.
As
future research work, we could point out that,
·
With all that is discussed in this study,
it is hoped that the results may contribute to future research and encourage
the improvement of reliability techniques.
As a continuation proposal of this work, it
is suggested to extend the proposed methodology to the TTR (time to repair)
statistical modeling.
·
Another approach to be considered in a future study, could be to perform
individual reliability analyzes for each critical production line equipment.
· In addition,
in a future study it is intended to analyze the impacts of predictive
maintenance of hibernation on the conservation of a plant's assets during the
off-season period, seeking to reduce maintenance costs and inappropriate
replacement of surviving parts from the previous crop.
REFERENCES
ALSYOUF,
I. (2009) Maintenance practices in Swedish industries: Survey results. International Journal of Production
Economics, v. 121, n. 1, p. 212-223.
DOI: https://doi.org/10.1016/j.ijpe.2009.05.005.
CARRETERO, J.; PEREZ, J. M.; GARCIA-CARBALLEIRA, F.; CALDERON, A.;
FERNADEZ, J.; GARCIA, J. D.; LOZANO, A.; CARDONA, L.; COTAINA, N.; PRETE, P.
(2003) Applying RCM in large scale systems: a case study with railway networks.
Reliability Engineering & System Safety, v. 82, n. 3, p. 257-273.
DOI: https://doi.org/10.1016/S0951-8320(03)00167-4.
CHOPRA,
A.; SACHDEVA, A.; BHARDWAJ, A. (2016) Productivity enhancement using reliability
centred maintenance in process industry. International Journal of Industrial and Systems Engineering, v. 23, n. 2, p. 155-165.
https://www.inderscienceonline.com/doi/abs/10.1504/IJISE.2016.076397
ELSAYED,
E. A. (1996) Reliability engineering. Massachusetts: Addison Wesley. European
Federation of National Maintenance Societies. http://www.efnms.eu/about-us/what-does-efnms-stand-for/
Access: 08/07/2018.
FARRERO,
J. C.; TARRÉS, L. G.; LOSILLA, C. B. (2002) Optimization of replacement stocks
using a maintenance programme derived from
reliability studies of production systems. Industrial
Management & Data Systems, v. 102, n. 4, p. 188-196. DOI: http://dx.doi.org/10.1108/02635570210423226.
FOGLIATTO, F. S.; RIBEIRO, J. L. D. (2009)
Confiabilidade e Manutenção Industrial. Rio de Janeiro: Elsevier.
GANDHARE, B. S.; AKARTE, M. M.; PATIL, P. P. (2018) Maintenance performancemeasurement–a case of thesugar industry, Journal of Quality in Maintenance Engineering, v. 24, n. 1, p. 79-100. DOI: https://doi.org/10.1108/JQME-07-2016-0031.
GIOLO, S. R.; COLOSIMO, E. A. (2006) Análise
de sobrevivência aplicada. Edgard Blucher.
HEO, J. H.; KIM, M.
K.; LYU, J. K. (2014) Implementation of reliability-centered maintenance for
transmission components using particle swarm optimization. International
Journal of Electrical Power & Energy Systems, v. 55, p. 238-245.
DOI: https://doi.org/10.1016/j.ijepes.2013.09.005.
KOMNINAKIS, D. (2017) Análise de
confiabilidade para formulação de estratégia de manutenção de equipamentos em
uma empresa da indústria alimentícia. 2017. 96 p. Dissertation (Master in
Production Engineering) ARARAQUARA: UNIARA. https://m.uniara.com.br/arquivos/file/ppg/engenharia-producao/producao-intelectual/dissertacoes/2017/denis-komninakis-diniz.pdf.
KOMNINAKIS, D.; PIRATELLI, C. L.; ACHCAR, J. A. (2018) Análise de
confiabilidade para formulação de estratégia de manutenção de equipamentos em
uma empresa da indústria alimentícia. Revista Produção Online, v.
18, n. 2, p. 560-592.
DOI: https://doi.org/10.14488/1676-1901.v18i2.2871.
KUEHNE Jr., M. (2004) Planejamento e acompanhamento
logístico-industrial como diferencial competitivo na cadeia de logística
integrada. Thesis (PhD in Industrial Engineering). Florianopolis: UFSC.
LAFRAIA, J. R. B. (2001) Manual de
Confiabilidade, Mantenabilidade e Disponibilidade. Rio de Janeiro: Qualitmark.
LAWLESS, J. F. (1982) Statistical models and
methods for lifetime data, Wiley series in probability and mathematical
statistics, Wiley & Sons.
LEWIS, E. E.
(1994) Introduction to Reliability Engineering. John Wiley & Sons.
MENGUE, D. C.; SELLITTO, M. A. (2013)
Estratégia de manutenção baseada em funções de confiabilidade para uma bomba
centrífuga petrolífera. Revista Produção Online, v. 13,
n. 2, p. 759-783. DOI: https://doi.org/10.14488/1676-1901.v13i2.1341.
MOBLEY,
R.K. (2002), An Introduction to Predictive Maintenance,
Butterworth-Heinemann, Second Edition.
MONTGOMERY,
D. C.; RUNGER, G. C.(2010) Applied statistics and probability for engineers,
5nd Edition, Wiley & Sons.
MOUBRAY,
J. (1997) Reliability-Centered Maintenance. Industrial
Press. New York.
RAPOSO, C. F. C. (2011) Overall
Equipment Effectiveness: aplicação em uma empresa do setor de bebidas do polo
industrial de Manaus. Revista Produção Online, v. 11,
n. 3, p. 648-667. DOI: https://doi.org/10.14488/1676-1901.v11i3.529.
RAUSAND,
M. (1998) Reliability centered maintenance. Reliability Engineering and
System Safety, v. 60, n. 2, p. 121-132. DOI: https://doi.org/10.1016/S0951-8320(98)83005-6.
RUSCHEL,
E.; SANTOS, E. A. P.; LOURES, E. F. R. (2017) Industrial maintenance
decision-making: A systematic literature review. Journal of
Manufacturing Systems, v. 45, p. 180-194. DOI: https://doi.org/10.1016/j.jmsy.2017.09.003.
SABOUHI,
H.; ABBASPOUR, A.; FOTUHI-FIRUZABAD, M.; DEHGHANIA, P. (2016) Reliability
modeling and availability analysis of combined cycle power plants. International Journal of Electrical Power & Energy Systems, v. 79, p. 108-119. DOI: https://doi.org/10.1016/j.ijepes.2016.01.007.
SELLITTO, M. A. (2007) Análise
estratégica da manutenção de uma linha de fabricação metal-mecânica baseada em
cálculos de confiabilidade de equipamentos. Revista GEPROS, n. 2, p. 97.
DOI: https://doi.org/10.15675/gepros.v0i2.157.
SELLITTO, M. A. (2005) Formulação estratégica da manutenção industrial
com base na confiabilidade dos equipamentos. Production, v. 15, n. 1, p.
44-59.
DOI: http://dx.doi.org/10.1590/S0103-65132005000100005.
TSAROUHAS,
P. (2018) Reliability, availability and maintainability (RAM) analysis for wine
packaging production line. International Journal of Quality &
Reliability Management, v. 35, n. 3, p. 821-842. DOI: https://doi.org/10.1111/j.1365-2621.2012.03073.x.
VISHNU, C.
R.; REGIKUMAR, V. (2016) Reliability based maintenance strategy selection in
process plants: a case study. Procedia technology, v. 25, p. 1080-1087. DOI: 10.1016/j.protcy.2016.08.211.
WAEYENBERGH,
G.; PINTELON, L. (2009) CIBOCOF: a framework for industrial maintenance concept
development. International Journal of Production
Economics, v. 121,
n. 2, p. 633-640. DOI: https://doi.org/10.1016/j.ijpe.2006.10.012.