Exploitation Efficiency System of Crane based on Risk Management

Janusz Szpytko

and Yorlandys Salgado Duarte

AGH University of Science and Technology, Krakow, Poland

Keywords: Risk Management, Stochastic Optimization, Maintenance Process.

Abstract: The subject of the paper is the exploitation efficiency system of overhead type cranes operating in critical

systems, results implementation the control risk management and maintenance scheduling processes. The

study case of the paper is a hot rolling mills system of a steel plant with critical overhead cranes operating

with hazard conditions and continuous operation. The model output is an optimal overhead cranes

maintenance scheduling distribution minimizing the production line risk stopped and the model input is a

digital database structure with historical information related with the operation, maintenance, logistics and

management process of the overhead cranes in the hot rolling mills plant. The transfer function is a stochastic

non-linear optimization model with bounded constraint that assess a risk global-system indicator based on

Monte Carlo simulations.

1 INTRODUCTION

Today’s industry has high levels of automation and

complexity, therefore, decompound the Lego system

into critical pieces simplifies the problem and gives

us the opportunity to focus on a specific process.

The process of degradation is inherent in the

technical system; consequently, control risk

management and maintenance scheduling processes

are increasingly relevant, and the human decision-

making process behind is a target to improve.

Human decision-making process behind of the

control risk management and maintenance

scheduling processes is the coordination of

components maintenance and/or replacement that

make up the system but maintaining risk holistic

and/or clustering objectives defined by the decision

makers.

Coordination of larges combinations, meaning

larges systems, can be a complex problem and

humanly dreadful to find a faster optimal solution.

Mathematically speaking is a nondeterministic

polynomial time NP-complete problem.

As we know, for the search of optimal solutions,

it is needed first, to model the system and its possible

operational scenarios, to make a coherent

coordination. Reason why, software, tools, robots,

https://orcid.org/0000-0001-7064-0183

https://orcid.org/0000-0002-5085-3170

platforms are needed to perform this coordination

duties efficiently.

In the field of control risk management, we found

the closed-loop engineering (CLE) framework

introduced by (Barari et al., 2009) as a well-stablish

approach for robust and coherent coordination.

Strong references of CLE implementation are

(Gholizadeh et al. 2020) for operational and tactical

decision-making levels to configure a coordinated

supply chain network, (Jerome et al. 2020) for

integration of production scheduling decisions within

a dynamic real-time and (Rui et al. 2020) to assess the

dynamic reliability of repairable closed-loop systems

with the consideration of uncertainties. All of them

are examples of platforms to support robust and

coherent coordination duties.

Following the same research line, in this paper, we

study the control risk management and maintenance

scheduling processes for overhead cranes operating in

a steel plant. An innovated exploitation efficiency

system of crane based on risk management is

proposed to simulate the same process performed in

the time real, but in this case, an optimization

algorithm chooses the best maintenance schedule

given the historical degradation data of the previous

process as a result of machine learning analysis, and

provides the feedback to the entity manager as a

Szpytko, J. and Duarte, Y.

Exploitation Efﬁciency System of Crane based on Risk Management.

DOI: 10.5220/0010123200240031

In Proceedings of the International Conference on Innovative Intelligent Industrial Production and Logistics (IN4PL 2020), pages 24-31

ISBN: 978-989-758-476-3

closed-loop control system. Among the cases cited,

the reference (Rui et al. 2020) maintenance decisions

oriented with a novel non-probabilistic reliability

assessment approach is an example close to the

proposal in this paper but in a different system.

Figure 1 describes the block diagram of the

engineering solution proposed, called in this paper as

Integrate Maintenance-e (IMe), e- referring to

electronic or digital.

Figure 1: Risk based Maintenance Management Efficiency

Control System.

In the presented model, the control risk

management effectivity depends on the maintenance

scheduling optimization. By nature, in the literature,

this process is an open problem because engineering

systems increase their complexity and variety every

single day and new researches are needed to fill the

gabs of the challenges.

Maintenance scheduling as a general problem,

can be decomposed essentially by hierarchical levels,

holistic objective (referring to global or integrated

strategies) or multi-objectives (referring to

decentralized strategies), and optimization criteria

cost, reliability, or hybrid approach. Examples are

cost-holistic (Akaria et al., 2019), cost-multi-

objectives sequential (Briskorn et al., 2019) and

holistic-reliability approach (Luo et al., 2019). For

any of the cases, the problem is to find the best

scheduled maintenance sequence of actions for each

component considered in the system.

Generally, the objectives and restrictions are not

well defined because depends on the individual

system requirements. However, as a consensus, the

optimization problem is defined as a multi-criteria

combinatorial problem of non-linear objective

functions with constrains (adding by us stochastics),

and the problem objective is to determine the timing

and sequence of the maintenance tasks periods of

each component analysed. Therefore, the variables x

in a maintenance scheduling problem is represented

by the start time of the maintenance tasks for all the

component considered.

Especially, the paper is focuses on defining the

exploitation efficiency system based on risk

management for overhead cranes under operation into

the steel plant, and in a specific scenario description

as an application example. The idea is to contribute

with an example of overhead cranes adaptation to the

digital industry and with a clear union of control risk

management with maintenance scheduling.

The motivation of the investigation starts with the

identification of organization issues in the dedicated

maintenance department of the steel plant, which is

focusing on cranes operation into the continuous

transportation process in the hot rolling mills system.

In this system, overhead cranes are critical devices,

because in case of failure or maintenance the

production line can stop.

The department have a risky situation also when a

scheduled maintenance of selected cranes (existing as

hot redundancy) is performed and at the same time an

unexpected fail of cranes in use in the system are

reported.

In practice, we consider a set of cranes in the

operation process of the plant. We learn about the

results of technical degradation of cranes under the

operation processes, implementation results of

dedicated to cranes maintenance focused procedures,

as well as the existing environmental conditions and

applied plant operation strategies.

The presented exploitation efficiency system

based on risk management for proper engineering

decision making and controls, considers the plant

operation strategies. The system platform helps also

to adapt crane maintenance processes to the existing

operation problems and events and available

resources, as well as unexpected critical events into

the hot mill system.

The IMe platform supports decision-making

processes aimed at minimizing the risk of the

operational safety of cranes and the risk of losing their

operational reliability, as a result of the degradation

of the structure and utility functions of devices and

the possibility of a combination and association of

events and failures resulting in a safety hazard under

operation processes.

In our case, maintenance-task distribution or

maintenance scheduling solution implemented is a

holistic-reliability approach.

The approach was selected holistic for an easy

interpretation of the maintenance impact on the

system by the entity manager (unique indicator), and

reliability, because overhead cranes in a steel plant

are critical devices working in a continuous process,

by construction, the system must be reliable.

In this paper, the approach selection is driven by

individual system requirements and the contribution

is strongly guided by the CLE framework (Barari et

al., 2009).

Conceptually, the proposed model considers two

layers (live and digital) guided by independent

Exploitation Efﬁciency System of Crane based on Risk Management

processes that are eventually are combined by the

platform.

Live layer presents manufacturing process in the

real time, representative by dedicated exploitation

data. Historical data from the crane operation process

is supplemented with current data obtained from the

transport process carried out by cranes, which are

available results with the use of sensors.

Digital layer based on Digital Twin (DT)

processes replaces human decisions making by risk

managements tools, incorporating optimization of the

decision making involved.

The first process concerns the registration of the

process of technical degradation of cranes

exploitation parameters and losses their functional

functions, because of the implementation of specific

transport processes.

The second process is oriented towards planning

service processes with the use of specific limited and

available resources, their correlation with planned

overhauls of a technological manufacture line and

their adaptation to the occurring unplanned events

and expectations. The process of planning service and

maintenance processes is accompanied by an

optimization process focused on the planned

efficiency of the production system and minimizing

possible production losses.

Such decision supporting model platform, based

on parallel real and digital processes, helps to design

optimal maintenance strategies dedicated to the

selected cranes, including type of the activities and

timetable, and needed resources. Is holistic, safety

and reliability oriented, includes quality and quantity

cranes representatives’ parameters for decision

making processes.

In practice, the input of the model is a database

structure created for maintenance purposes based on

historical degradation data of all the system

components, previous planned process, system

structure, etc., collected by SCADA (Supervisory

Control And Data Acquisition) and SAP (Systems,

Applications and Products in Data Processing)

systems, in fact, available historical information

related to the maintenance scheduling management

process.

The output of the model in our case is to provide

to the entity manager a faster and optimal online

decision-making process as a closed-loop control

system (CLCS).

Once the maintenance scheduling management

process is optimal (supporting by live and digital

layers), as a result, the holistic operational efficiency

in the plant increases. In the following sections, the

model is fully described using a specific scenario.

The document is organized as follows; firstly, the

mathematical formulation of the optimization

problem is presented with the constraint set and the

flow diagram, as well as all the equations and

assumptions of the proposed model organized in

subsections. Following the same idea, in next sections

the scenario (parametrization) and methods (solution)

are described, discussed, and validated. Finally, some

important conclusions are drawn to highlight

potential outcomes in this research area.

2 IME-PLATFORM:

MATHEMATICAL

MODELLING

The proposed model aim is to minimize the expected

value of the convolution function, between the

overhead cranes loading system capacity distribution

function of the steel plant impacted by maintenance

scheduling and the necessary load capacity

distribution function of the production line. The

model is defined below:

{

}

,1 , ,1

,1 ,1 , 1 ,

min [ ]

where :

(,)

(, , )

( , , , , ) with

subject to:

0 Simulation Window ( )

iikik i

i i ik ik

RfXY

XftLC

LC f t TTM TDM x x TTM

TTM TTM TTM TDM

≠

→

→=

<< − + +



(1)

The stochastic non-linear optimization model

with bounded constraints proposed for the overhead

cranes’ maintenance scheduling problem solution in

the steel plant present only continuous variables x =

TTM

i,1

(time to maintenance) and is defined in the

model constraint intervals. The independent variable

of the objective function to be optimized x = x

, x

…, x

NMi

depends on the quantity of maintenance task

to be coordinated for each overhead crane.

The optimization variables are only the start times

for the first maintenance of each overhead crane

TTM

i,1

. Once TTM

i,1

is established, the remaining

TTM

i,k≠1

are calculated adding the corresponding

maintenance intervals, which are invariable and

depends on the operation time between two

consecutive check-ups.

Figure 2 shows the conceptual flow diagram

implemented to solve the problem. The flow diagram

is linear and only has two conditioning moments, first

one to guarantee the simulations error criterion, and

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics

second one to guarantee the best solution of all the

scheduling proposals evaluated.

Figure 2: Flow optimization model.

The problem is solved as follows: the

optimization algorithm proposes a set of TTM

i,1

and

Monte Carlo simulations are performed to

determine N

values of risk (r

, r

, …, r

). The Risk

mean E[R] and variance V[R] are determined from (r

, …, r

) and the error criterion is checked. If the

desired error is not achieved, N

is augmented and the

Monte Carlo simulations are repeated for the same set

of TTM

i,1

. When the desired error is achieved the

process is repeated for another set of TTM

i,1

. This is

done several times (1, 2, 3, …, N) determining

decreasing values of Risk mean E[R

], E[R

], …,

E[R

]. The set of TTM

i,1

leading to the lowest Risk

mean (E[R

]) is the solution.

The optimization model construction formalized

in this section is structured in two steps. First one, the

production-line-capacity and overhead-cranes-

capacity stochastic mathematical models for a steel

plant are defined. The model used for the overhead

cranes has two reliability states and considers random

faults intrinsic to these systems, faults repair times

and operation standards. Second one, the Monte Carlo

simulation model used to estimate the risk indicator

(Capacity Loss) is formalized.

2.1 Production Line Capacity

Modelling

In a continuous production process, always the

devices target function, at all the hierarchical levels is

to be available. Redundancies for critical devices are

crucial in continuous process to obtain high reliability

performance, therefore, as an example the steel plant

have more overhead cranes than needed.

In this paper based on a steel plant scenario, we

define the production line capacity dependent on the

total overhead cranes loading capacity of the system.

In the proposal model, we call this parameter η -

Production line efficiency because is an indicator

between [0, 1] and measures the relation between

minimum availability of overhead cranes required to

secure the production line and total overhead cranes

loading capacity of the system.

In addition, for maintenances purposes also, but

not related with the overhead cranes system, the steel

plant need to stop the production line N

SPL

times,

therefore two additional parameters STD

(stooped

time duration) and TBS

(time between stooped) are

introduced into the model. As a result, the production

line capacity is defined as follows:

111

11 11

()

0if

ikk

mm m m

kk kk

LC t TBS STD

Yft

TBS STD t TBS STD

−

===

−

== ==









+≤<+







 

(2)

where t is time, θ is a set of parameters, in this

case θ = {η,

, TBS

, STD

} and m = 1, 2, 3, …,

SPL

2.2 Overhead Cranes Modelling

Overhead cranes operation is continuous, eventually

fails and is repairable. This random behavior can be

described from distribution functions fitted to

historical degradation data. Considering the operation

effectiveness of the overhead crane, in this paper, we

fix that the system and its components have two states

z = 0, 1 and between them transition rates are defined

depending of the distribution function selected in the

simulation approach. The probability of moving from

one state to another depends on the failure or repair

rate of each overhead crane.

In the two-state model, the overhead cranes are

considered fully available (z = 1) or totally

unavailable (z = 0). The stochastic loading capacity

at the time instant t of an overhead cranes i is

determined by the TTF

i,k

(time to failure), TTR

i,k

(time

to repair) and

(nominal loading capacity). The

parameters allow to simulate with (3) the behavior of

generating k-th independent random numbers

from distribution functions fitted to historical

degradation data. Therefore, the model proposed to

simulate the stochastics overhead cranes loading

capacity is defined below:

Exploitation Efﬁciency System of Crane based on Risk Management

,, ,,

11 11

()

0if

iikik

mm m m

ik ik ik ik

kk kk

LC t TTF TTR

LC f t

TTF TTR t TTF TTR

−

== ==









+≤<+







 

(3)

where t is time, θ is a set of parameters, now θ =

{

, TTF

, TTR

i,k

} and m = 1, 2, 3, …, N

RNi

knowing that N

RNi

depends on the Simulation Window

used in the optimization model. The k-th independent

random numbers generated from distribution

functions fitted to historical degradation data

guarantees the follows restriction:

Simulation Window

ik ik

TTF TTR

+≥



On the other hand, one of the factors that affects

the overhead crane loading capacity, is not stochastic

and is not considered a random phenomenon, it is

type maintenance process, and in this paper, we

assume independent from the previous one during the

simulation. The maintenance is contemplated within

the strategies of a steel plant because it guarantees

cranes life cycle. Maintenance is the activity designed

to prevent failures in the production process and in

this way reduce the risks of unexpected stops due to

system failures. In a steel plant, to perform some

maintenance tasks it is necessary that the overhead

crane does not work, and this causes Capacity Loss in

the steel plant. Due to this reason, it is advisable that

this maintenance task be carried out at the time of the

year where the least frequency of system potential

failure exists, so that equilibrium and adequate

environment are secured in the steel plant. To

consider this effect, in this paper the parameters

TTM

i,k

(start time to maintenance) and TDM

i,k

(time

duration maintenance) are introduced in the equation

(4), then we combine the equation (3) and (4) using

junction symbol & representing the AND logic, as we

show below:

,, ,,

11 11

()

0if

iikik

nn n n

ik ik ik ik

kk kk

LC t TTM TDM

LC f t

TTM TDM t TTM TDM

−

== ==









+≤<+







 

(4)

where t is time, θ is a set of parameters, now θ =

{

, TTM

i,k

, TDM

i,k

} and n = 1, 2, 3, …, N

MTi

; and

as a combination consequence of both process LC

& LC

knowing that the junction symbol & is

used for the AND logic. As a modelling result, the LC

variable consider both process, degradation (D) and

maintenance (M).

Once we know the LC

for each overhead crane,

we use reliability block diagrams to compose the

system loading capacity. A technical complex system

structure can be reduced in a series-parallel reliability

block diagram, and the steel company is not the

exception. Usually, the steel companies have huge

warehouses, and inside overhead cranes are installed,

therefore the series-parallel configuration of the

cranes follows the structure of the warehouse. Based

on the configuration of the warehouse is possible built

the system block diagram, and as a result, the system

simulation process is consequent with the relation

between cranes.

In order to consider the block diagram structure of

the system, we propose a simple rule in this paper.

Following the generic series-parallel structure, when

two or more overhead cranes are in series (Crane

Crane

& … Crane

) the junction symbol & is used

for the AND logic, while the symbol || is used for the

OR logic when the overhead cranes are in parallel

(Crane

|| Crane

|| … Crane

), therefore, during the

simulation process of the system when two or more

overhead cranes are in series, if one crane fail, all the

chain of cranes in series stop, overwise, when two or

more cranes are in parallel, if one crane fail, the

redundancy system is still working.

As a conclusion, we simulate independently the

for each overhead crane i, then we combine all

the N-series cranes in each M-parallel chain of cranes

using the junction symbol &, and then we aggregate

all the equivalent M-parallel chain of cranes using the

junction symbol || to obtain the system loading

capacity X. Below, we define the general notation for

the overhead cranes system loading capacity:

()( )

|| &

XLC

(5)

2.3 Risk Indicator Modelling

The risk function denoted as R can be generated with

the sum of X + Y random, independent, and non-

negative variables. By definition, the product of R(s)

= P(s)Q(s) is defined with the generating function

()

Ps ps

∞



of X and the generating function

()

Qs qs

∞



of Y. Consequently, the generating

function of R(s) is generally defined by the

convolution formula:

kjkj

rpq

−



(6)

where p

and q

are the generated sequence from

P(s) and Q(s) respectively.

In this investigation, X is the overhead cranes

system loading capacity distribution function affected

by the maintenance scheduling defined in (5), and Y

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics

is the production line capacity defined in (2). The risk

function is denoted in this investigation as R and is

defined below as a convolution product between (5)

and (2):

0if

tt tt

YX XY



−<







≥





(7)

where t = 1, 2, 3, …, T (Simulation Window).

The expected value of the risk function E[R] is

defined in this paper as Capacity Loss. In this work,

to estimate E[R] the Monte Carlo simulation method

is used. The convergence process is fluctuating in this

method. However, the error level decreases when the

number of samples increases, according to the law of

large numbers. In this method it is not practical to run

a simulation with many samples, because more

calculation time is required. Therefore, it is necessary

to balance the required precision and the calculation

time with a stop criterion. This criterion guarantees

that the simulation continues, until the risk indicator

has the precision specified for the simulation. The

parameter used as stopping criterion in the method is

the coefficient of variation β defined below.

[] []

[] [] [] [] []

[]

CapacityLoss ER ER ER ER ER

ER N N

σσ

=± =± =±

⋅

(8)

3 IME-PLATFORM:

MATHEMATICAL

PARAMETERIZING

The system analyzed have 33 overhead cranes with

loading capacity between 2-80. Depending on

operation time and loading capacity, each crane has

weekly, every two weeks or monthly inspection

frequency as we describe in the Table 1.

Table 1: Inspection frequency relation.

Capacity

(tons)

Inspection

frequency

TTM

(hours)

TDM

(hours)

50T – 120T Weekly 168 3

32/8T – 20T

Every 2

weeks

336 3

5T-8T Monthly 672 3

The inspection frequency are maintenance tasks,

therefore in the model are considered as TTM

i,k

and

TDM

i,k

,, where k-th is the number of inspections for

each overhead crane depending on the simulation

window. In the historical degradation data case, the

fitted distribution function for each crane is a result of

the fitting-selection decision making flow diagram

from previous work.

In order to parameterize the degradation process

of the system, the Table 2 summarize all the finals

distributions selected for each overhead crane.

Table 2: Degradation distribution parameters.

Crane

Failure time

distribution

Repair time

distribution

807 Exponential

µ = 743.80

Lognormal

µ = 1.07; σ = 0.93

808 Weibull

a = 1796.93; b = 0.60

Generalized Pareto

k = -1.05; σ =

12.56;

θ = 0

809 Gamma

a = 0.54; b = 2607.04

Inverse Gaussian

µ = 6.82; λ = 4.70

810 Exponential

µ = 9184

Exponential

µ = 3.25

870 Birnbaum Saunders

β = 377.02; γ = 2.59

Loglogistic

µ = 0.98; σ = 0.48

871 Weibull

b = 0.64; a = 1060.19

Inverse Gaussian

μ = 2.30; λ = 4.43

1011 Exponential

µ = 3254.86

Exponential

µ = 8.60

1010 Inverse Gaussian

µ = 476.41; λ = 79.40

Inverse Gaussian

µ = 5.9; λ = 1.87

872 Exponential

µ = 2396

Exponential

µ = 6.63

873 Exponential

µ = 5571.4

Exponential

µ = 4.53

874 Exponential

µ = 1189.2

Exponential

µ = 72.5

879 NaN NaN

1001 Weibull

b = 0.82; a = 458.66

Lognormal

µ = 1.44; σ = 1.14

1000 Burr

α = 2318.3; c = 0.75; k =

3.15

Inverse Gaussian

µ = 10.44; λ = 4.59

1002 Exponential

µ = 3285.9

Exponential

µ = 12.95

1003 NaN NaN

1004 NaN NaN

1005 Weibull

b = 0.70; a = 429.64

Inverse Gaussian

µ = 65.65; λ = 2.93

1006 Exponential

µ = 2534.4

Exponential

µ = 5.07

1007 Exponential

µ = 1125.7

Exponential

µ = 2.6

1008 NaN NaN

1009 NaN NaN

1016 NaN NaN

1021 NaN NaN

502 Exponential

µ = 1478.2

Exponential

µ = 4.38

981 Birnbaum Saunders

β = 508.68; γ = 3.13

Inverse Gaussian

µ = 16.82; λ = 2.11

983 Weibull

b = 0.62; a = 611.51

Inverse Gaussian

µ = 8.52; λ = 3.53

Exploitation Efﬁciency System of Crane based on Risk Management

Table 2: Degradation distribution parameters (cont.).

Crane

Failure time

distribution

Repair time

distribution

985 Generalized Extreme

Value

k = 0.8; σ = 1176.1; µ =

903.95

Inverse Gaussian

µ = 72.79; λ = 2.69

804 Exponential

µ = 4008.8

Exponential

µ = 3.59

813 Burr

α = 1027.3; c = 0.78; k =

2.92

Inverse Gaussian

µ = 22.66; λ = 2.73

815 Nakagami

µ = 0.28; Ω = 2852.2

Inverse Gaussian

µ = 3.88; λ = 2.03

812 Exponential

µ = 8564.6

Exponential

µ = 6.44

811 Exponential

µ = 8803.5

Exponential

µ = 4.18

Note: NaN means non failures registered.

Once we know all the parameters related with the

overhead cranes system capacity, the next step is the

production line capacity. Two essentialise

information, the monthly STD is 12; 10; 16 and 12

hours every week respectively, therefore STB

between them are 168 hours; and the efficiency

indicator is 85% in the simulated scenario.

In the case of the simulation parameters, the

simulation window is one year (8760 hours) and we

assume robust expected value estimation (Capacity

Loss indicator) by Monte Carlo simulation when ε =

0.01.

4 IME-PLATFORM:

MATHEMATICAL SOLVING

The model is fully implemented in MATLAB

(R2019b), a multi-paradigm numerical computing

environment and proprietary programming language

developed by MathWorks and can be running in any

personal computer.

As we describe above, the model has a stochastic

non-linear objective function with bounded

constraints. In order to solve this specific problem,

PSO algorithmic was used because Capacity Loss risk

indicator (objective function value given the

maintenance scheduling) is the results of a

convolution by Monte Carlo simulation, therefore we

do not know the objective function derivate and

Newton's, Lagrange, quasi-Newton or Sequential

Quadratic Programming traditional methods cannot

be used.

Knowing the features of the objective function,

during the model implementation three possible well-

stablished algorithms to solve derivative free

problems were found and tested: GA (genetic

algorithm), PSO (particle swarm optimization), and

Nelder-Mead modified (NMm).

GA as a global searching algorithm, in large

search regions needs numerous evaluations in the

objective function to find the minimum.

NM, by definition is a searching algorithm

without restriction, but during the implementation

was possible bound the independent variables of the

objective function to adapt the algorithm to the

problem (NMm). NM has a limitation related with the

number of independent variables. Independently of

the objective function, when the number of

independent variables is more than ten, the algorithm

rarely finds the global optimum if the initialization of

the search is not accurate.

Given the previous statements, PSO a local

searching algorithm, is the option selected to be used

because behaves better in this particular problem,

finds the solution with less evaluations in the

objective function, and as a consequence, the time

needed to solve the problem is lower than GA. In

addition, the number of independent variables is not

a limitation for this algorithm.

PSO is fully applicable to this problem. PSO

algorithmic used in this investigation is based on the

algorithm described in (Kennedy et al., 1995), using

modifications suggested in (Mezura-Montes et al.,

2011) and in (Pedersen, 2010). PSO algorithm iterates

until it reaches a stopping criterion, in this case, when

the relative change in the best objective function

value is less than 1.0000e-06 (Function Tolerance

described in the diagram flow).

Once the optimization algorithm used on the

solution and the full parameterization are described,

the results of the proposed optimization model given

the steel plant scenario is shown in Figure 3.

Figure 3: Convergence process of the optimization

algorithm.

Figure 3 is the convergence process of the

optimization algorithm and shows how the Capacity

Loss decrease when the maintenance scheduling

change. Proper planned maintenance scheduling

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics

process improve the operational efficiency in the steel

plant, and the CLCS guaranty the expected results.

The exploitation efficiency system based on risk

management is valuable for the entity manager

because he/she can decide according to standards risk

level, what would be the best moment in the year to

perform the maintenance process in the system.

Simulation-based approaches are powerful for

modeling stochastic processes with complex

functions, but the time to simulate these processes can

be a limitation with the current computing power. In

our case the average time duration of ten consecutive

simulations performed with an i5 5250U 1.6 GHz

CPU for the described parametrization is [(1.955813

± 0.072131)·Ns·E] seconds.

5 CONCLUSIONS

The paper describes with a parameterized scenario,

how the exploitation efficiency system based on risk

assessment find the optimal overhead cranes

maintenance scheduling in the steel plant.

Experimental results show that presented closed-loop

control model can help to organize the maintenance

scheduling strategy in the steel plant. The paper

solves the assessing risks problem of transportation

process in the steel plant through the simulation-

based approach which considers the relationship

between random factors (historical degradation data

fitted by machine learning framework) during the

production process and maintenance scheduling

process (planned process, making-decision

framework).

The presented model has the advantage of

minimum set of data needed for robust decision

making, but two fissures are in place, the model is a

local focused problem solution (unique), means, we

do not have any comparative reference to assess the

model, just validations by steps and study cases, and

the time because we use simulation-based approach.

The local solution is well accepted by the steel

plant and futures steps of the investigation will

recover the results of the application in practice, but

still remain open the generalization of the proposal in

others system with similar orientation problems.

The presented model opens the way to extensive

simulations under various scenarios and conditions,

with the possibility to be updated in real-time, to

detect anomalies, to control systems and to conduct

accurate diagnostics and prognostics of cranes into

selected scenarios.

ACKNOWLEDGEMENTS

The work has been financially supported by the

Polish Ministry of Science and Higher Education.

REFERENCES

Barari, A., and Pop-Iliev, R. (2009). Reducing rigidity by

implementing closed-loop engineering in adaptable

design and manufacturing systems.

Journal of

Manufacturing Systems,

28, 47-54.

Dirk Briskorn, Lennart Zey. Interference aware scheduling

of triple-crossover-cranes.

Journal of Scheduling.

https://doi.org/10.1007/s10951-019-00634-6.

Hadi Gholizadeh, Hamed Fazlollahtabar. Robust

optimization and modified genetic algorithm for a

closed loop green supply chain under uncertainty: Case

study in melting industry.

Computers & Industrial

Engineering

, 147 (2020) 106653.

Islam Akaria, Leah Epstein. An optimal online algorithm

for scheduling with general machine cost functions.

Journal of Scheduling. https://doi.org/10.1007/s10951-

019-00629-3.

Jerome E.J. Remigio, Christopher L.E. Swartz. Production

scheduling in dynamic real-time optimization with

closed-loop prediction.

Journal of Process Control, 89

(2020) 95–107.

Kennedy, J., and R. Eberhart. Particle Swarm Optimization.

Proceedings of the IEEE International Conference on

Neural Networks

. Perth, Australia, 1995, pp. 1942–

1945.

Mezura-Montes, E., and C. A. Coello Coello. Constraint-

handling in nature-inspired numerical optimization:

Past, present and future.

Swarm and Evolutionary

Computation

. 2011, pp. 173–194.

Pedersen, M. E. Good Parameters for Particle Swarm

Optimization. Luxembourg: Hvass Laboratories, 2010.

Rui He, Guoming Chen, Xiaoyu Shen, Shengyu Jiang,

Guoxing Chen. Reliability assessment of repairable

closed-loop process systems under uncertainties.

ISA

Transactions

, 104 (2020) 222–232.

Wenchang Luo, Yao Xu, Weitian Tong, Guohui Lin.

Single-machine scheduling with job-dependent

machine deterioration.

Journal of Scheduling (2019)

22:691–707. https://doi.org/10.1007/s10951-019-

00622-w.

Exploitation Efﬁciency System of Crane based on Risk Management