Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Variable control tool in MATLAB for energy transformation processes

To cite this article: Y Cardenas et al 2020 J. Phys.: Conf. Ser. 1708 012035

View the article online for updates and enhancements.

This content was downloaded from IP address 186.116.67.150 on 15/02/2021 at 20:49

VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

Variable control tool in MATLAB for energy transformation

processes

Y Cardenas1, G E Carrillo2, A Alviz3, and G Carrillo4

1
Universidad de la Costa, Barranquilla, Colombia
2
Universidad Tecnológica de Bolívar, Cartagena de Indias, Colombia
3
Universidad Industrial de Santander, Bucaramanga, Colombia
4
Corporación Politécnico de la Costa Atlántica, Barranquilla, Colombia

E-mail: [email protected]

Abstract. During the stages of transformation of energy in a process, exercise control over the
variables that intervene in it, improve its performance, and identify undesirable conditions in
these. Thus, this study is developed as a graphical interface to implement a methodology for
controlling variables of energy conversion processes, such as internal combustion engines. The
control tool developed in MATLAB variables is based on multivariate statistics. The methods
for developing this tool of Graphic User Interface is based on the statistics of principal
component analysis and failure statistics such as T ! Hotelling and the Q statistic that allows the
control of anomalies presented in the operation's behavior. About the methodology, first, the
input data are normalized, achieving standardization of the observation matrix vs. variables, then
the spectral decomposition of the normalized data is performed, reaching the generation of the
matrix of auto-values, allowing the age of the projection space of the data. With this based and
delimited, it is possible to establish the ranges of observation of the mentioned statisticians. The
result obtained from this research corresponds to software that allows the constant observation
and analysis of the behavior of each variable of the generation engine. It describes the upper
limit, lower limit, arithmetic mean, principal components, graphics of the statistics, and detects
the failures in real times.

1. Introduction
Today, the concepts and methods of statistical process control have become a fundamental aspect for
monitoring operations in the manufacturing and energy-generating industries [1]; itis worth mentioning
that both the concepts and the methods of the internal combustion engine (ICE) are complementary to
those of the control of the automatic feedback process. The control tools allow monitoring the
performance of a process over time to verify that it is maintained in a "statistical control state" [2]. Such
a control state exists if the process or product variables remain close to their desired values or in the
"normal" operating range [3]. Based on this idea, what is expected through the variable's operational
control is to identify the root causes of the failure, achieve improvements in the process chain, avoid
corrective maintenance, and apply bases of preventive care.
Conventional multivariate statistical process control schemes focus on monitoring the stability of the
arithmetical mean of the process, limiting the analysis of the "weighting" or valuables variables that
directly affect the process to a value higher than 80 % of the total incidence [4]. Thus, detecting and
diagnosing the variables that cause failures has become one of the most widely used methods worldwide
to identify abnormalities that impact any decision-making point. However, the growing complexity of
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

operating electricity generation plants means that these activities are carried out with more exceptional
care by the personnel in charge, requiring computer platforms that facilitate this work [5]. One of these
platforms has been the so-called graphic user interfaces (GUI), which allows visual interaction
composed of Image, and graphic objects of the process with the user or operator role [6]. The use of
GUI has been developed as a control information tool in the industrial sectors. Usually, the software
systems have been built on platforms based on events or functional parameters that allow the variability
of the analysis independence of the study's settings. The software-driven by technical parameters is
contained by a GUI that enables the user to use digital buttons as a tool to t variables [7]. Statistical
studies show that 70% of industrial events are caused by human error [8,9], requiring the incorporation
of control techniques for industrial automation and methods of fault detection and diagnosis (FDD) in
different industrial environments; an example FDD development is the petrochemical industries [10],
transport [11] and energy specifically in generation engines [12] among others.
Although the conventional control is based on robust statistical and graphical control settlements, it
suffers from a lack of applicability in data-rich environments, typical of modern processes such as power
generation supported by ICE's [13]. To monitor the state of the process from the principal component
analysis (PCA), different statistics are used to build control charts to monitor the process's condition.
Thus, it is possible to follow the process in a univariate way and see the reality in the threshold. The
most widely used statistics at the industrial correspond T ! of hotelling and statistical Q [14]. Thus,
multivariate statistical projection methods, such as PCA, where both statistics are used, are used to
reduce the surveillance space's dimensionality by projecting the information of the original variables in
low dimensional subspaces defined by a few latent variables.
The main contribution of this research is the development of a tool that provides a constant analysis
of the variables of operation of an internal combustion engine, in which transformation of the chemical
energy of a fuel such as natural gas, mechanical work of rotation of a shaft and then electricity, in a
generator system occurs. This tool is developed under the MATLAB environment with the fundamentals
of multivariate analysis of principal components.

2. Methodology
This section presents the mathematical description of the principal component analysis used for the GUI
design based on ICE's failure behaviour. This methodology allows for constant monitoring of the
variables involved in energy transformation [15].

2.1. Statistical components

It is considered a data set consisting of observation variables and observations for each variable, which
is arranged in a matrix arrangement X" ∈ ℝ 𝐧∙𝐦 as shown in the agreement, Equation (1) [15].

𝓍&& 𝓍&! … 𝓍&'

𝓍!& 𝓍!! … 𝓍!'
X" = + ⋮ ⋮ … ⋮ /. (1)
𝓍(& 𝓍(! … 𝓍('

The matrix X" should be normalized or standardized in such a way as to allow analysis without the
tendency to the units of each m. From this matrix, it is possible to calculate the covariance matrix as
defined by Equation (2) [16].

𝟏
S = 𝐧*𝟏 X 𝐓 ∙ X, S ∈ ℝ 𝐦∙𝐦 . (2)

Spectral decomposition of S can be obtained if their values or self-values and their respective
individual vectors or self-vectors are found. You can find the eigenvalues, λ, , of S solving the Equation
(3) [15].

2
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

det(λI − S) = 0 → λ& , λ!…, λ' , (3)

I is the identity matrix. Self-assessments must be organized in such a way that the conditional
expressed in the condition of Equation (4).

λ& ≥ λ! ≥ ⋯ λ' . (4)

The spectral decomposition is then performed obtaining Λ and V where the auto-value matrix Λ is
obtained with the arrangement as shown by the expressed in the condition of Equation (5) [15]:

λ& 0 … 0
0 λ& … 0
Λ=+ / , 𝚲 ∈ ℝ 𝐦∙𝐦 , (5)
⋮ ⋮ ⋱ ⋮
0 0 … λ'

and the autovector matrix expressed in the vector of Equation (6).

V = [ν& ν! ⋯ ν' ], V ∈ ℝ '∙' , (6)

the decomposition of S then, the disintegration of is done as shown by the Equation (7) [17].

S = V ∙ Λ ∙ V 1. (7)

Projection y = V 1 ∙ x of an observation vector x =∈ ℝ ' m converts the observation space into a set
of uncorrelated variables corresponding to the elements of y. The variance of the i = ésimo element of
y is equal to i = ésimo self-worth in the matrix Λ. Where S s invertible (since it is a symmetric matrix),
it is possible to apply the definition shown by the Equation (8) [17].

z = Λ*&/! ∙ V 1 ∙ 𝓍 (8)

So, the statistician T ! hotelling is given by Equation (9) [15].

T ! = z 1 ∙ z. (9)

The projection of the observations, contained in X, in a smaller dimensional space will be included
in the scoring matrix described by the Equation (10) [15].

T = X ∙ P ∈ ℝ (∙3 , (10)

where the first self-values are contained in a matrix P ∈ ℝ '∙3 and the projection T back to the space
of observation m = dimensional is, Equation (11).

R= T ∙ P1
X (11)

The residual matrix is defined as follows, as shown in the Equation (12) [15].

R,
E=X−X (12)

where the initial data set is expressed by the selected main components and the noise space.

3
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

3. Results
This section presents the results based on the design of the control tool based on the multivariate analysis
under the PCA statistics described in the methodology. The algorithm or steps that were for the
development of this GUI tool are broken down; additionally, it is shown how the normalized values of
failures of the ICE and number of principal components according to their weighting in the spectral
space are projected in the designed tool. These results allowed a constant monitoring of the variables
involved in the process of energy transformation.

3.1. Description of the design algorithm

Using MATLAB, interface Figure 1 was designed for fault detection based on the algorithm used in this
study. Below is a flowchart that shows the general content of the GUI developed. Figure 1 describes
step by step the content of the GUI. Initially, the tool is designed with a presentation of the authors, also
contains an installer for those users who do not have MATLAB on their computers. Then the platform
is designed to receive the information from the matrix described in the methodology; this matrix contains
observations vs variables, if it includes the requirements described in the tutorial, conditions such as that
the excel has no special features such as NAN, among others, it continues to the analysis of the behavior
of variables individually, Next, the process of quantifying and characterizing the main components is
carried out, identifying the anomalies with the T ! Hotelling and the Q statistic, finally, the detection and
identification of the failure to monitor the variables involved in energy transformation.

Figure 1. Graphic user interface flowchart.

3.2. Graphic user interface

The program FDD-PCA v1.0 is a graphic user interface that operators in charge of ICE's control can
use; the interface allows multivariate statistical analysis based on PCA; de GUI's presentation is shown
in Figure 2. Figure 2 shows the general presentation of the platform; this presentation is contained by
general graphics of what the software presents. It should be noted that the two statistics are shown with
what the data processing is done.
Figure 3 shows the behaviour of the statisticians T ! Hotelling and the Q statistic, the measurement
of the variability of the 9898 measurements performed.
The observations made in the observations from 1 to 1000 indicate that the level of operation with
respect to the following observations is low, this level of normal operation, is consistent with the regular
behavior of the variables in control status. However, they present instantaneous disturbances in very
short periods of time, which does not symbolize alarms in the operators because the behavior returns

4
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

quickly to its average. Additionally, there is the Q statistic, which shows the residual error behavior that
have the variables in space T ! . Figure 4 shows the individual behavior of each variable, it is presented
the ranges of upper limit, lower limit and arithmetic mean of the behavior of the variables, facilitating
the operator the visualization of the real behavior. The first a vectores column in the autovector matrix
V indicate the main direction of the space in which the variables move with their respective auto-values
Λ or scale factor. The criterion of choice of a corresponds to the explanation of at least 95% of the
variability contained in the matrix that initially entered.

Figure 2. Presentation of the graphic user interface.

Figure 3. Installer, tutorial, and authors’ section.

In the Figure 5. the principal components are quantified and characterized according to the input
matrix's nature. According to the gobblers, the case study of the internal combustion engine presents 30
members with their respective weights. Finally, the GUI becomes a useful tool in detecting the failure
in the ICE process. The exciting thing about the tool is that it allows us to visualize the moment of
observation in which the system presents loss.

5
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

Figure 4. Individual analysis of the variables.

Figure 5. Characterization of the principal components.

4. Conclusions
It is a GUI tool that manages to detect and identify failures and the origin of these during the ICE
operation. This educational tool can be used at the industrial level by operators as an instrument of
association with policies of proper use of energy resources and sustainability policies, since early
detection of faults will help to prevent the system from operating outside acceptable or permissible
operating conditions, preventing fuel consumption and increased emissions of gases such as CO2, and
generating gases such as NOx, SOx, etc., created during the combustion process in gasoline, natural gas,
diesel, gas turbine, boiler-steam turbine assembly engines.
The inclusion of the GUI in energy generation processes addresses one of the lines (preventive and
corrective maintenance) in energy management systems in industries, managing to facilitate observation
and monitoring of the behavior of each of the fundamental components for the operation of the energy
transformation machine.
The results obtained in the analysis show that for the study of the energy transformation machine or
an internal combustion engine, the variable of greater significance corresponds to the Voltage of the
cylinder 8 (x8), with a percentage of 38% throughout the 9898 observations in units of time. The
algorithm developed in the GUI step by step shows the precise time the anomaly begins that corresponds
to the comment 1000, which means that the monster was presented before the operators' report since the
operators show abnormality in the analysis 7600.

6
VII International Week of Science, Technology, and Innovation (7th IWSTI) IOP Publishing
Journal of Physics: Conference Series 1708 (2020) 012035 doi:10.1088/1742-6596/1708/1/012035

References
[1] Belussi L, Danza L, Salamone F, Meroni I, Galli S, Svaldi S 2017 Integrated smart system for an energy
audit: Methodology and application Energy Procedia 11(140) 231–239
[2] Inoue M, Urata K 2017 Performance evaluation and reinforcement in interconnected passive system IFAC-
Papers On Line 50(1) 9999–10004
[3] Marofi R 2014 Selection of maintenance strategies based on AHP and TOPSIS techniques Nat. Sci. 12(10)
163–168
[4] Perera L, Machado M, Manguinho D, Valland A 2016 System failures of offshore gas turbine engines in
maintenance perspective IFAC-PapersOn Line 49(28) 280–285
[5] Xie C, Grechanik Q, Fu M 2013 A GUI differentiator IEEE International Conference on Software
Maintenance (Edmonton: IEEE)
[6] Bouraoui A, Gharbi I 2019 Model-driven engineering of accessible and multi-platform graphical user
interfaces by parameterized model transformations Sci. Comput. Program. 172(63) 101-113
[7] Phannachitta P 2020 On an optimal analogy-based software effort estimation Inf. Softw. Technol. 125(10)
106-116
[8] Morag I, Chemweno P, Pintelon L, Sheikhalishahi M 2018 Identifying the causes of human error in
maintenance work in developing countries. J. Ind. Ergon. 68(12) 222–230
[9] Jang I, Kim A, Jung W, Seong P H 2014 An empirical study on the human error recovery failure probability
when using soft controls in NPP advanced MCRs Ann. Nucl. Energy 73(12) 373–381
[10] Marais H, Van G, Uren K 2019 The merits of exergy-based fault detection in petrochemical processes J.
Process Control 74(1) 110–119
[11] Zheng T, Tan R, Li Y, Yang B, Shi L, Zhou T 2016 Fault diagnosis of internal combustion engine valve
clearance: The survey of the-state-of-the-art Proc. World Congr. Intell. Control Autom. 13(12) 2614–2619
[12] Vasu J,. Deb A K, Mukhopadhyay S 2015 MVEM-based fault diagnosis of automotive engines using
Dempster-Shafer theory and multiple hypotheses testing IEEE Trans. Syst. Man, Cybern. Syst. 45(7) 977–
989
[13] Campos J, Cardenas Y, Valencia G 2017 Trends in Failure Studies of Generation Engines based on
Statistical Models from 2007 to 2017 JESTER 11(3) 163–167
[14] K Lu, Jin Y, Chen Y, Yang Y, Hou L, Zhang Z, Li Z, Fu C 2019 Review for order reduction based on
proper orthogonal decomposition and outlooks of applications in Mech's mechanical systems. Syst. Signal
Process. 123(12) 264–297
[15] Chiang L, Russell E, Braatz R 2010 Fault diagnosis in chemical processes using Fisher discriminant
analysis, discriminant partial least squares, and principal component analysis Chemometrics and Intelligent
Laboratory Systems 12 243–252
[16] Boutellaa E, Kerdjidj O, Ghanem K 2019 Covariance matrix based fall detection from multiple wearable
sensors J. Biomed. Inform 94(12) 103-119
[17] Cardenas Y 2019 Fallas en Bujias para Motores de Generación a Gas (Colombia: Universidad del
Atlantico)