Article
Article
Article
Computational and
Experimental Methods
in Mechanical
Engineering
Proceedings of ICCEMME 2021
Smart Innovation, Systems and Technologies
Volume 239
Series Editors
Robert J. Howlett, Bournemouth University and KES International,
Shoreham-by-sea, UK
Lakhmi C. Jain, KES International, Shoreham-by-Sea, UK
The Smart Innovation, Systems and Technologies book series encompasses the
topics of knowledge, intelligence, innovation and sustainability. The aim of the
series is to make available a platform for the publication of books on all aspects
of single and multi-disciplinary research on these themes in order to make the latest
results available in a readily-accessible form. Volumes on interdisciplinary research
combining two or more of these areas is particularly sought.
The series covers systems and paradigms that employ knowledge and intelligence
in a broad sense. Its scope is systems having embedded knowledge and intelligence,
which may be applied to the solution of world problems in industry, the environment
and the community. It also focusses on the knowledge-transfer methodologies and
innovation strategies employed to make this happen effectively. The combination of
intelligent systems tools and a broad range of applications introduces a need for a
synergy of disciplines from science, technology, business and the humanities. The
series will include conference proceedings, edited collections, monographs, hand-
books, reference books, and other relevant types of book in areas of science and
technology where smart systems and technologies can offer innovative solutions.
High quality content is an essential feature for all book proposals accepted for the
series. It is expected that editors of all accepted volumes will ensure that
contributions are subjected to an appropriate level of reviewing process and adhere
to KES quality principles.
Indexed by SCOPUS, EI Compendex, INSPEC, WTI Frankfurt eG, zbMATH,
Japanese Science and Technology Agency (JST), SCImago, DBLP.
All books published in the series are submitted for consideration in Web of Science.
Computational
and Experimental Methods
in Mechanical Engineering
Proceedings of ICCEMME 2021
Editors
Veeredhi Vasudeva Rao Adepu Kumaraswamy
Department of Mechanical & Industrial Department of Mechanical Engineering
Engineering Defence Institute of Advanced Technology
College of Science, Engineering Pune, India
and Technology University of South Africa
Pretoria, South Africa Ambuj Saxena
Department of Mechanical Engineering
Sahil Kalra G. L. Bajaj Institute of Technology &
Department of Mechanical Engineering Management
Indian Institute of Technology Jammu Greater Noida, India
Jammu, India
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface
v
vi Preface
We are thankful to all sponsored agencies who gave us their cooperation and
funding support.
We are thankful to our management and the director of G L Bajaj Institute of Tech-
nology and Management, Greater Noida, Uttar Pradesh, India, for their continuous
source of inspiration and valuable support. We are thankful to all the members of the
organizing committee for their contribution in organizing the conference. Last but
not least, we thank Springer for its professional assistance, particularly, Mr. Aninda
Bose who supported this publication.
vii
viii Contents
xiii
xiv Editors and Contributors
Sahil Kalra was born in Kaithal, a small city in Haryana. He completed his B.Tech.
in Mechanical Engineering from Kurukshetra University in 2011. During his B.Tech.
degree, he received a full tuition fee waiver from the university. In his first attempt,
he qualified GATE exam with 96 percentile and joined M.Tech. in the Department
of Mechanical Engineering, NIT Jalandhar. Based on the performance, he was given
an opportunity to serve as Assistant Professor in the same institute. However, after
serving for one semester, he moved to IIT Kanpur as a full-time Ph.D. student.
There he worked on a research project sponsored by Space Application Centre,
Ahmedabad, Indian Space Research and Organization (ISRO). He worked jointly
with the ISRO scientists and developed a reconfigurable antenna for space applica-
tions. He defended his Ph.D. degree in May 2, 2019. From his Ph.D. research, he
has published four reputed International Journal articles, two patents and six confer-
ences, and symposiums. He is recipient of two international and two national awards
at different conferences and symposiums. He has been offered a prestigious fellow-
ship, namely Faculty in Science and Engineering, at the University of Groningen
(QS ranking 65) in the Netherlands (Europe). Currently, he is serving as Assistant
Professor in the Department of Mechanical Engineering, IIT Jammu. His research
interests are system dynamics and control, mechatronics, robotics, vibration control,
modal analysis, smart materials and smart structures. He is passionate about the algo-
rithm development for different finite element method-based engineering problems
and its solution using using Abaqus and MATLAB.
Contributors
29.1 Introduction
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 323
V. V. Rao et al. (eds.), Computational and Experimental Methods in Mechanical
Engineering, Smart Innovation, Systems and Technologies 239,
https://doi.org/10.1007/978-981-16-2857-3_30
324 J. Awatramani et al.
decrease the number of yearly visits. The safety issues of elevators have transformed
into a global elevator issue. Therefore, elevator systems require proper maintenance
to ensure safety and reliability. Although, there are various obstacles faced during
the actual maintenance of elevator systems such as inappropriate maintenance tech-
niques, inappropriate upgrading of elevator instruments, high maintenance costs, and
much more. Developing preventive and predictive maintenance arrangements will
be the next phase for revamping the security of elevators, which will further escalate
the life of elevator systems and decrease the cost of service. Preventive mainte-
nance performs systematic inspections of assets and executes routine maintenance
to prevent unexpected equipment downtime or failure. Preventive maintenance can
decrease the frequency of equipment failure. Though this maintenance procedure
needs a definite cost, there is a possibility that the cost of the whole elevator system
might increase if over-maintenance of equipment is being done. If preventive main-
tenance activities are not useful enough, the rate of equipment failure will be a high
rise, and post-maintenance costs of the whole system might also increase.
Predictive maintenance lets you estimate the time-to-failure of a machine. So we
can start preventive maintenance and save time and assets from any big issue. Predic-
tive maintenance strategies are nowadays being implemented by elevator production
service companies. This implementation comes into use by evaluating the remaining
life of the elements that are accountable for faults and remotely tracking faults in
elevator systems. Fault detection and diagnosis are needed by elevators to perform a
healthy operation.
The traditional predictive maintenance machine learning models are based on
feature engineering, which is the manual construction of the right features using
domain expertise and similar methods. This usually makes these models hard to
reuse since feature engineering is specific to the problem scenario and the available
data, which vary from one place to the other.
The most attractive part of applying deep learning in the predictive maintenance
domain is the fact that these networks can automatically extract the right features
from the data, eliminating the need for manual feature engineering. Deep learning
approach provides better results due to the new deep features extracted from the
dataset compared with the existing features.
This paper explores the elevator system’s optimal maintenance policy to reduce
the average maintenance cost. The rest of the work has been structured as follows:
Sect. 29.2 provides the in-depth literature review, Sect. 29.3 showcases gap analysis,
Sect. 29.4 represents the parameter analysis, Sect. 29.5 represents the implementation
of random forest classifier in our model, Sect. 29.6 represents the experimental results
on the same and Sect. 29.7 showcases the conclusion along with the future scope.
29 Investigating Strategies and Parameters to Predict … 325
Several search engines were used for online library search: IEEE Xplore Digital
Library, Springer Link Online Library, Elsevier ScienceDirect and Google Scholar.
Most of the authors have implemented deep learning strategies in their models in
order to obtain results [1, 2, 5–7]. An effective fault diagnosis and fault prognosis
has been observed. Along with which, implementation of decision trees and random
forest algorithm has also been observed in order to compare the results. The use of
deep learning has helped the authors to extract new deep features from the dataset as
well. These new deep features showcased better accuracy in terms of fault diagnosis.
Mishra et al. [1, 2] designed a generic deep encoder model to automatically calcu-
late highly explanatory deep features from the elevator data. Random forest algorithm
was applied to determine faults based on statistical attributes to compare results.
Strategy with new extracted deep attributes provided great accuracy in determining
the faults. In this paper [5], the authors have proposed a conventional framework
of intelligent predictive maintenance systems for elevator service consisting of fault
prognosis, fault diagnosis, feature extraction, pre-processing signal, IoT and Internet
of Service. Gilabert et al. [6] studied predictive maintenance as a result of uncrit-
ical machinery. Continuous monitoring along with diagnosis has been done using
neural networks, whereas, machine instruments were being observed using vibration
systems and applied using Bayesian networks. It also showed the work carried out
under the MINICON project that tends to develop cost-effective integrated SPUs.
In this paper [7], a consolidated neural-network based on decision support structure
for predictive maintenance of rotating equipment. It comprises a vibration-based
deterioration database by observing rolling element bearings, developing an ANN
model to estimate the life percentile, along with the failure times of roller bearings
and cost matrix designing, and probabilistic substitution model that tends to enhance
the expected cost per time.
Few papers have showcased the implementation of machine learning algorithms
including SVM, genetic algorithms [3, 8, 9]. They focused mainly on reducing the
mean cost of the maintenance of an elevator and finding out the most optimal strategy
to estimate the time of maintenance. These papers consisted of static datasets only
and detection of faults at an early stage.
In this paper [3], the model is combined with a numerical example that is solved
by a genetic algorithm with the goal of reducing the mean cost of maintenance. The
best equipment maintenance plan of action is found by verifying the practicality and
validity of the model. In Langone et al. [8], LS-SVM at an early stage of fault detection
was used. Initially, on sensor data coming from the VFFS machine, an unsupervised
approach, KSC came into play. After that, the NAR model was illustrated, i.e., a
supervised learning methodology, LS-SVM framework. It showed that LS-SVM
can successfully assess, predict mechanical conditions based upon sensor data and
attain higher performance than basic methods. In this paper [9], a discussion on
Multivariate Relevance Vector Machines that examines direct and indirect procedures
326 J. Awatramani et al.
for forecasting everyday evapotranspiration was being done. The developed models
are based on static datasets.
Few miscellaneous studies were mainly based on the research on the mechanical
strategies, which included examination of hydraulic oil, MCSA and algorithms and
penalty functions to make a degradation and maintenance model [10–13]. These
papers were research-based and showcased the impact of predictive maintenance on
elevator systems.
Xu et al. [10] drew an outline about the role of predictive maintenance in the
epoch of big data and techniques of data-directed fault diagnostics and fault prog-
nostics along with its future extent. In this paper [11], a dynamic predictive mainte-
nance policy for a multi-component approach. It also provides an outline of literature
on predictive and condition-based/vanilla maintenance strategies. Flores et al. [12]
designed Motor Current Signature Analysis (MCSA) is an identification method-
ology in which the mechanical structure acts as a transducer, determining short
torque variations produced within the mechanical structure. MCSA has been used
for determining the condition of various mechanical components such as bearings,
motor fan, rotor unbalance and gearboxes, providing information in regards with fault
centralization. In this paper [13], various methods have been performed to inspect
the percentage of metals and corrosion; sulfur levels in the hydraulic fluid depict the
maintenance intervals of an elevator system. The study of zinc, phosphorus, chlorine
and calcium verifies that the oil that was used in one elevator is different from the oil
used in the other two. Additionally, it releases significant wear to all the elevators as
a result of the working environment conditions. This study of the hydraulic fluid can
be included in the testing process, which inspects the safety of elevators (Fig. 29.1).
Fig. 29.1 Timeline evolution of techniques to predict the health of elevator systems
29 Investigating Strategies and Parameters to Predict … 327
In [3], the mathematical explanation was presented, which reduced the mean main-
tenance cost, but the discussion about the relationship between the two equipment
was not discussed. Although the analytical model used in [5] was pretty advanced,
still, there is a need for an efficient data-mining approach and analysis model. In this
study, the main focus of this work was on diagnostics and prognostics techniques,
managing the tasks that are scheduled and with the restricted data, predicting the
machine RUL. But, challenges faced by the authors are to improve the efficacy of
the model [10]. In [11], to improve the efficacy of maintenance. The work focused
more on the component dependencies rather than system dependencies. To look for
various techniques in order to model stochastic dependence. The work in [7] show-
cased the optimization of maintenance of task scheduling. Challenges were faced
while recording the readings of bearing when replaced in group. Cost matrix, along
with margin life distribution, should also be expanded.
Elevators work on the principle of the see-saw. The central module of an elevator
is the elevator car, in which the commuter or load is carried. The elevator car has a
supporting assembly attached to it, that assembly is attached to a traction rope. The
vertical motion of the elevator is guided by a pair of rails. The elevator is raised and
lowered with the help of traction steel rope and counterweight.
The rope is looped around a pulley, which is known as a driving pulley [14]. The
driving pulley is attached to an electric machine. They together form a system known
as a traction system. Traction system runs by the elevator’s control system present
in the elevator, which allows its motion.
There are many mechanical components present in an elevator system, these
components are in direct relation with the performance and fidelity of the elevator.
The traction system is the driving force of an elevator. The pulley present in the
traction system consists of ball bearing and this wears off due to fretting friction
caused between the pulley and the steel ropes [15]. Counter-weight balances the
elevator car, which puts constant stress on the traction cable, the stress increases
at the time of vertical movement of the elevator system. The motor present in the
traction system produces short torque vibration, which helps determine the longevity
of the elevator system. Distance traveled by the elevator is in direct relation with the
elevator as the travels more distance the components wear off. Other components
such as motor fans, gearboxes, humidity also affect the health of the elevator systems
[12].
Some parameters have more impact on the health of the elevator than others.
Ball-bearing present in the driving pulley is responsible for friction-free movement
of the elevator system. This component of the traction system wears off as there is
328 J. Awatramani et al.
constant stress present from the traction steel ropes. Fitting friction among the ropes
causes curtailment of the cross-sectional area of the metal, which naturally leads to
resistance on the pulley. As a result, the bearing wears off. Distance traveled by an
elevator is a clear indicator as to how much an elevator is used. As a consequence, it
helps determine the health, wear and condition of the elevator across a timeline [16].
Torque vibration refers to the vibration produced by the mechanical motor present
in the traction system, this attribute determines the health of the motor.
The dataset used [4] has been recorded from a variety of IoT sensors used in the
elevator industry. It contains the evening usage of an elevator between 16:30 and
23:30. The dataset consists of mainly three attributes: ‘Ball-Bearing’, ‘Vibration’
and ‘Humidity’. Dataset provides operational data in the form of time series. It
showcases the functional life of an elevator system, which is observed to be roughly
27 years. On analysis, it has been observed in Fig. 29.2 that there is a constant
decline in revolutions (rpm). A perfectly functional ball bearing has 93 rpm, and it
deteriorates as low as 13 rpm in a span of 27 years. The average life of an elevator is
20–25 years. Hence, by observing the graph, it is determined that at a value between
35 and 45 rpm, the elevator is most likely to cause a failure.
In Fig. 29.3, it is observed from the available statistics that in the initial 5–6 years
of the elevator system, minor changes occur. The elevator system starts experiencing
fluctuation in vibration after 7–10 years of use through the occurrence is less. The
major increase in the occurrence of vibration is observed after 15–20 years of use.
As traction motors are the source of short torque vibration, the health of the motor
can be determined by the intensity of the vibrations.
Classification is one of the biggest parts of machine learning and random forest
classifier holds the top of the classifier hierarchy [1, 2]. It is basically a bagging
technique in which there is a collection of a large number of decision trees. Decision
trees have two important properties of low bias and high variance. Random forest
classifier creates an array of decision trees that are created from randomly generated
subsets of training data. Then, votes from the different decision trees are combined
to determine the final category of the test object. Random forests can also apply the
weight concept, which takes into account the magnitude of influence of decision tree
outcome. Trees with low error values have high weight values and vice versa. This
increases the impact of trees that have low error rates. The following parameters have
been used in the study:
• Gini Impurity: calculates the purity of the split and it is better than entropy, as it
is computationally efficient. The range of gini impurity lies between 0 and 1.
• N_estimators: 20 (total number of trees)
• max_depth: 3 (depth of an individual tree in forest)
• min_samples_split: 2 (minimum samples required to split branch of the tree)
• min_sampels_leaf: 1 (minimum samples required to be at leaf node)
Random forest classifier assigns a random value to the root node sample and
classification condition which in our case is “samples = 44408” and “ball-bearing <=
40.045”. According to the classification condition, the range of samples is classified
into three distinct states: Good (0), Fair (1) and Poor (2). This process continues until
all the samples are classified successfully.
In Fig. 29.4, a random forest classifier selects a randomly selected condition value,
which is close to the midpoint which in our observation is “ball-bearing <= 40.045”.
So, all the sample points with ball bearing less or equal to 40.045 are categorized as
class 2. In the second level of the tree, the leaf node signifies the number of samples
that are categorized as class 2. The non-leaf node on the second level of the tree
represents the remaining samples, which are further divided into two subtrees based
330 J. Awatramani et al.
on the randomly selected condition value selected by the random forest classifier
“ball-bearing <= 60.661”. This process is repeated until all the samples are classified.
An average accuracy of about 91.50% (ranging 91% ± 2%) was achieved using a
random forest classifier (Fig. 29.5).
After the implementation of the random forest classifier, it was observed that the
ball-bearing feature came out to be of more significance than the vibration feature.
This was due to the frequent fluctuations seen in the readings of vibration in our
dataset.
29.7 Conclusion
References
1. Mishra, K., Huhtala, K.: Condition Monitoring of Elevator Systems Using Deep Neural
Network, pp. 381–387 (2020). https://doi.org/10.5220/0009348803810387
2. Mishra, K.M., Huhtala, K.J.: Fault detection of elevator systems using multilayer perceptron
neural network. In: 2019 24th IEEE International Conference on Emerging Technologies and
Factory Automation (ETFA), Zaragoza, Spain, 2019, pp. 904–909. https://doi.org/10.1109/
etfa.2019.8869230
3. Liu, H.J., Wu, J.X.: Research on preventive maintenance strategy of elevator equipment. Open
J. Soc. Sci. 6, 165–174 (2018). https://doi.org/10.4236/jss.2018.61012
4. Axenie, C., Bortoli, S.: Predictive maintenance dataset (2020). https://zenodo.org/record/365
3909#.X2cDzz_itPY
5. Kesheng, W., Guohong, D., Lanzhong, G.: Intelligent predictive maintenance (IPdM) for
elevator service—through CPS. IOTaS Data Min. (2016). https://doi.org/10.2991/iwama-16.
2016.1
6. Gilabert, E., Arnaiz, A.: Intelligent automation systems for predictive maintenance: a case
study. Robot. Comput.-Integr. Manuf. 22(5–6), 543–549 (2006). ISSN 0736-5845. https://doi.
org/10.1016/j.rcim.2005.12.010
7. Wu, S., Gebraeel, N., Lawley, M.A., Yih, Y.: A neural network integrated decision support
system for condition-based optimal predictive maintenance policy. IEEE Trans. Syst. Man
332 J. Awatramani et al.
A G
Amardeep, 9, 367, 469 Gandhi, Sonal, 397
Ameer, Aatif, 509 Govil, Vinamra Kumar, 227
Ansari, M. A., 259 Gupta, Anuj, 429
Ansari, N. A., 519 Gupta, Deepa, 207
Awatramani, Jasmine, 323 Gupta, Shorya, 299
Gupta, Shradha, 197
Gupta, Tarun Kumar, 77
B Gurjar, Siddharth Singh, 445
Baghel, Navneet Singh, 19
Bahri, Harshit, 115, 177
Bansal, Ashish, 19 H
Bansal, Chetan, 411 Hanief, Mohammad, 241
Bansal, Mohit, 275 Hasteer, Nitasha, 323
Bhati, Aditya Kumar, 445
Bhatia, Siddhant, 469
I
Islam, Anas, 357, 481
C
Chaturvedi, Rishabh, 357, 387, 419, 459
Chauhan, Pankaj Kumar, 43
J
Chauhan, Sandeep, 19, 77, 411
Jain, Shruti, 489
Chauhan, Shailendra Singh, 445
Jha, Mayank Shekhar, 131
Chauhan, Vinit, 33
Joshi, Shivani, 397
Chauhan, Yash, 445
Chhonkar, Ankit, 19
Choudhary, Shilpa, 377
K
Kalra, Sahil, 1, 131
D Kaura, Shivendra, 275
Datta, Priyanka, 275, 287 Khan, Sabah, 43
Dohare, Deepti, 333 Kirti, 489
Kumar, Aditya, 9
Kumar, Deepak, 1, 131
F Kumar, Harshit, 367
Farozan, Shama, 105 Kumar, Ishu, 469
© The Editor(s) (if applicable) and The Author(s), under exclusive license 539
to Springer Nature Singapore Pte Ltd. 2022
V. V. Rao et al. (eds.), Computational and Experimental Methods in Mechanical
Engineering, Smart Innovation, Systems and Technologies 239,
https://doi.org/10.1007/978-981-16-2857-3
540 Author Index
T
R Tanveer, Mohd. Qamar, 349
Rai, Amrita, 313 Tiwari, Divakar, 509
Rani, Pooja, 1, 245 Tiwari, Rajesh, 189
Rawat, Ritwick, 9 Tiwari, Vikas, 509
Rohilla, Rajesh, 287 Tomar, Mihir, 445
Tripathi, Ashish, 499
Tripathi, Pragati, 259
S
Saraswat, Manish, 155, 481
Satyapal, Harendra Kumar, 105 V
Saxena, Alok, 509 Vashist, Prem Chand, 499
Author Index 541