Page 1 of 8 - Cover Page Submission ID trn:oid:::30744:108023393

submission
My Files

My Files

University

Document Details

Submission ID

trn:oid:::30744:108023393 4 Pages

Submission Date 2,406 Words

Aug 12, 2025, 11:46 AM GMT+5:30

13,971 Characters

Download Date

Aug 12, 2025, 11:47 AM GMT+5:30

File Name

[email protected]

File Size

394.8 KB

Page 1 of 8 - Cover Page Submission ID trn:oid:::30744:108023393

 Page 2 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

7% Overall Similarity
The combined total of all matches, including overlapping sources, for each database.

Filtered from the Report

Bibliography

Quoted Text

Match Groups Top Sources

14 Not Cited or Quoted 7% 4% Internet sources

Matches with neither in-text citation nor quotation marks
4% Publications
0 Missing Quotations 0% 5% Submitted works (Student Papers)
Matches that are still very similar to source material

0 Missing Citation 0%
Matches that have quotation marks, but no in-text citation

0 Cited and Quoted 0%

Matches with in-text citation present, but no quotation marks

Integrity Flags
0 Integrity Flags for Review
Our system's algorithms look deeply at a document for any inconsistencies that
No suspicious text manipulations found. would set it apart from a normal submission. If we notice something strange, we flag
it for you to review.

A Flag is not necessarily an indicator of a problem. However, we'd recommend you

focus your attention there for further review.

Page 2 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

 Page 3 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

Match Groups Top Sources

14 Not Cited or Quoted 7% 4% Internet sources

Matches with neither in-text citation nor quotation marks
4% Publications
0 Missing Quotations 0% 5% Submitted works (Student Papers)
Matches that are still very similar to source material

0 Missing Citation 0%
Matches that have quotation marks, but no in-text citation

0 Cited and Quoted 0%

Matches with in-text citation present, but no quotation marks

Top Sources
The sources with the highest number of matches within the submission. Overlapping sources will not be displayed.

1 Internet

iaeme.com 2%

2 Publication

Oguntade, Tomiwa Isaac. "Investigation of Surfactant-Enhanced Oil Recovery in … <1%

3 Submitted works

University of New York in Tirana on 2021-04-30 <1%

4 Publication

Nina Prakash, Paul Gasper, Francois Usseglio-Viretta. "ExpertSegmentation: Seg… <1%

5 Internet

medium.com <1%

6 Internet

python.hotexamples.com <1%

7 Publication

Arthur S. de Sena, Pedro H. J. Nardelli, Daniel B. da Costa, Ugo S. Dias, Petar Popo… <1%

8 Submitted works

Frederick University on 2023-06-10 <1%

9 Publication

Pedregal, D.J.. "State space models for condition monitoring: a case study", Reliab… <1%

10 Submitted works

University of Portsmouth on 2025-05-16 <1%

Page 3 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

 Page 4 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

11 Internet

ksme.or.kr <1%

12 Publication

"Key Digital Trends Shaping the Future of Information and Management Science"… <1%

Page 4 of 8 - Integrity Overview Submission ID trn:oid:::30744:108023393

 Page 5 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

Smart Diagnosis Of Rotating Machinery

Dhaaniya N Deporal K Mrs Kayalvizhi K R
1
1 Electronics and Communication Electronics and Communication Assistant professor
Engineering Engineering Electronics and Communication
St.Joseph’s College Of Engineering St.Joseph’s College Of Engineering Engineering
[email protected] [email protected] St.Joseph’s College Of Engineering
[email protected]

vibration data generated according to ISO 10816. The

9 Abstract— Condition monitoring is essential in order to objective is to do automated motor health state classification
increase the reliability of industrial motors. Even though using a strong and interpretable model, XGBoost, and train it
standards like ISO 10816 have been developed, not all using well-structured input features such as vibration velocity
systems providing vibration data have automated and the machine class.
interpretation means. In this regard, this work presents
a novel machine learning method, based on artificial Apart from model building, the Snowflake platform
vibration data in accordance with ISO 10816, to classify performed cloud-based data storage, transformation, and
motor health states. This dataset contains 2000 samples control. Snowflake’s scalable compute architecture, SQL
annotated with Vrms, type, and health labels (A–D), support, and compatibility, and integration with ML
whose vibration severity levels are described. An workflows make it a great platform for processing structured
XGBoost model was developed to characterize motor data, doing massive-scale queries, and preparing data for
health as Good, Satisfactory, Unsatisfactory, and training. We can offload data wrangling and aggregations to
Unacceptable. Machine class and Vrms were identified Snowflake we get high performance, low latency, and
as the important features of the model, which is of high reusability of the data pipeline.
accuracy.
The proposed architecture blends adherence to an ISO
Keywords— vibration monitoring, ISO 10816, XGBoost, standard with real-time cloud-integrated analysis to provide
machine learning, motor health classification, predictive an easily scalable architecture for Predictive Maintenance.
maintenance Utilizing the power of both machine learning and
contemporary data platforms, Snowflake, this work provides
I. INTRODUCTION a bridge between theoretical vibration standards and practical
motor health assessment in an automated fashion within the
Electric motors are foundational to industrial operations and industrial setting.
power machinery, such as compressors, pumps, and turbines.
8 Ensuring their continuous and reliable operation is essential
for minimizing downtime, reducing maintenance costs, and II. LITERATURE SURVEY
maintaining production efficiency. This reduces the power
10 losses due to heat generation and sound generation. Vibration condition monitoring is a fairly well-established
method for early fault detection in rotating machines such as
This pass leads to minimization of the losses of power of motors, compressors and gearboxes. Many researchers have
thermal (and sound) origin, consequently growth of the level investigated the modelling of the physical vibration
of the vibrational activity of the engine indicates describing phenomenon and the application of machine learning for
its technical condition [1]. Since antiquity, engineers have automatic fault classification.
recognized the technical state of a car by the kind of noise it
makes. The method of vibration monitoring is grounded in Meli et al. [1] proposed an original modelling approach to
this principle. Accordingly, in this work, we tend to study and take a coupling between rotating machinery and flexible
analyze the review on motor vibration and monitoring foundations into account. Their work showed the importance
techniques for vibration monitoring. of modelling of machine– foundation dynamics in the real
The ISO 10816 has been established in order to standardize world where the vibration energy propagated and was
the interpretation of vibration levels by taking into measured on the stationary component. This observation is
consideration the Vrms, machine class, and mounting type, consistent with the objectives of the current hereinafter
and establishing the severity of vibration of a machine. So, described study, in which vibration velocity Vrms measured
these conditions, Good (A) to Unacceptable (D), really serve on non-rotating structures constitutes the primary feature set.
as a model for when you should know when you need to
maintain that asset. While it has the potential to be highly The standard, referred to as ISO 10816 [2], offers an
effective and accurate, many industrial settings do not have internationally accepted methodology for the assessment of
existing automated systems to correspond live, real-time machine vibration level. It classifies the machine health into
sensor data to these labels; therefore, as is, the standard is not four zones (A: Good, B: Fair, C: Poor, D: Failure) according
very useful in real-world use. to two levels of the Vrms and machine categories. This work
explicitly takes ISO 10816 into account for the annotation of
To meet this challenge, the current work introduces a
machine learning-based classifier developed on the synthetic

Page 5 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

 Page 6 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

training data and its predictions are interpretable and

actionable in industrial applications.

In the early days, Graybeal's seminal work [3] researched the

sources of vibration in the electric machines, reasoned causes
including the magnetic asymmetry, rotor eccentricity, and
bearing wear. These factors have a direct impact on the time-
domain vibration signals discussed in this study, emphasizing
the importance of the early work to the development of
contemporary feature extraction methods.

Li and Xu [4] described the reliability evaluation for a system

subject to instantaneous and degradation failures. Their
framework was at the system level, but the idea of modelling
multiclass gradual failure was similar to the four ISO-based
condition zones used in this paper. Their results support the A. Data Preprocessing in Snowflake ML Container
notion that successful fault diagnosis has to account for both
After selecting the model, we moved on to pre-processing,
unique and changing fault patterns.
training, and evaluation in the Snowflake ML container
environment. This setup allows for scalable data handling, in-
Yavanarani [5] presented work that implemented a 3D memory processing, and isolated machine learning
spectral method to investigate complex vibration signals execution. We retrieved the structured dataset that meets the
occurring in UGVs. The use of frequency-domain features by ISO 10816 guidelines from the Snowflake table
them further highlights the significance of better signal vibration_data using the Snowpark API and converted it into
processing techniques for increased classification accuracy. a pandas DataFrame.
Although this work concentrated on the time-domain features We transformed the target variable ZONE, which represents
including RMS, crest factor, and kurtosis, other extensions to machine health conditions (Categories A–D), into an integer
the frequency and time-frequency domains can be conducted. label column, Zone_Encoded, using the LabelEncoder. The
feature matrix X does not include the original or encoded
Afanaseva [6], a vibrational activity diagnostics of internal target labels, and we used the encoded label as the response
combustion engine and using a similarity theory and variable y. We performed a stratified train-test split using an
dimensional analysis to determine physical dependencies 80-20 ratio with random_state = 42, ensuring a consistent
describing the relation between parameters of the system and class distribution across the subsets.
vibrations. Their approach focused on a physics-informed
B. Label Encoding
diagnosis which is an alternative to our data-driven machine
learning approach in this study. The categorical label ZONE represents ISO 10816 health
categories (A to D). We used LabelEncoder to change the
Widodo and Yang [7] employed support vector machines string labels into numbers (Zone_Encoded). This change was
(SVM) for fault diagnostics of mechanical systems and important for compatibility with scikit-learn and XGBoost,
showed robust performance for bearing and gear faults. Their which need numerical target labels.
work can be considered as precursor work of applying Considering the possibility of class imbalance, a common
supervised learning algorithms to condition monitoring tasks. issue in fault detection datasets, the training pipeline included
Nevertheless, their method involved manual feature selection stratified sampling during the train-test split phase.
and was not automated for model benchmarking, a challenge Stratification made sure that all classes, particularly those less
that we filled by adopting AutoML-based model competition represented (like rare fault conditions), were adequately
approach using AutoGluon. represented in both the training and testing datasets. This
approach helped prevent the model from favoring the
majority classes
C. Hyperparameter Optimization and Model Training
III. METHODOLOGY
The chosen classifier, XGBoost, was trained and optimized
This section details the procedure taken for construction of in the Snowflake ML container. We used a manual grid
3 the motor health classification framework. It covers data search strategy to adjust the hyperparameters. This focused
preprocessing in Snowflake and model training in the on the number of estimators, maximum tree depth, and
Snowflake ML container using XGBoost. All stages were learning rate. We looked at the following configurations:
2 designed to have scalability, high precision and meet the ISO • n_estimators: {100, 300, 500}
10816 standard. • max_depth: {3, 5, 6}
• learning_rate: {0.1, 0.05, 0.01}
5 We evaluated each combination using three-fold cross-
validation with the cross_val_score function from the scikit-
6 learn library. The best configuration, n_estimators = 300,
max_ depth = 5, and learning_rate = 0.05, scored an average

Page 6 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

 Page 7 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

accuracy of 99.88% across the validation folds. We retrained measures were displayed on screen and we used the trained
this model on the entire training data and tested it on the XGBoost model to recognize our input. The input
holdout set. identification made with the following features: RMS = 2.5,
PEAK = 5.2, CRESTFACTOR = 2.00, KURTOSIS = 3.0,
and SKEWNESS=0.5 was considered properly classified. It
D. Snowflake ML Container as Execution Environment
could be shown that the system could correctly judge
All modeling tasks, including data transformation, splitting, vibration severity class Zone B shown in Fig. 6. This
training, and validation, were done in the Snowflake ML demonstrates the good performance of the model for online
container. This offered a company a secure and a scalable zone classification, and hints its potential use for industrial
environment for their Python-based machine learning that IoT and predictive maintenance applications.
included widely used packages like pandas, scikit-learn and
XGBoost. Using containers follows modern MLOps
principles, ensuring reproducibility, dependency isolation,
and smooth integration with Snowflake-hosted data.

E. ISO 10816 Alignment

A key strength of the framework is its compliance with ISO
10816, which is an international standard for evaluating
machine vibration severity. By classifying machine
conditions A (good) through D (unacceptable), the model
outputs are predictive and interpretable within a globally
recognized maintenance framework. This compliance allows
for deployment in industrial environments where ISO
compliance is required.

IV. RESULTS C. Confusion Matrix Analysis

7 In this section, we present the performance of the proposed The model exhibits a strong classification power with very
motor health classification scheme. The best-performing little errors. For instance, in Zone B 100 samples, 1 of them
12 XGBoost model was evaluated based on accuracy, precision, was misclassified as Zone C, and in Zone D, just 1 mistake
11 recall, and Cross-Validation scores. The results verify the was made. The other zones were well predicted.
effectiveness of the proposed model and its adherence to ISO
10816-based health classification.

A. XGBoost Model Training

A multi-class objective with optimized hyperparameters—
4 250 estimators, a maximum depth of 4, and a learning rate of
0.07—was used to train the final XGBoost model. When
tested on the test dataset, the model's classification accuracy
was 99.5%, as Fig illustrates. The robustness of the model in
capturing vibration zone patterns across the four ISO 10816-
based categories is demonstrated by this high accuracy. In
order to balance model complexity and generalization and
maintain high predictive performance with little overfitting,
parameter tuning was essential.

Train data Test data

Accuracy 98.5 98.8

B. Real-Time Prediction Output

To verify whether it was capable of prediction, we applied a
synthetic test input which was that the machine’s vibration

Page 7 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

 Page 8 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393

D. Classification Metrics analysisof dimensions,” in 2017 IEEE II International Conference on

Control in Technical Systems (CTS), Oct 2017, pp. 93–95.
[7] M. Widodo and B. Yang, “Support vector machine in machine condition
monitoring and fault diagnosis,” Mechanical Systems and Signal
Processing, vol. 21, no. 6, pp. 2560–2574, Aug. 2007.
[8] “IEE Colloquium on 'Advanced Vibration Measurements, Techniques
and Instrumentation for the Early Prediction of Failure' (Digest
No.105)," IEE Colloquium on Advanced Vibration Measurements,
Techniques and Instrumentation for the Early Prediction of Failure,
London, UK, 1992, pp. 0_1-.
[9] D. S. Broomhead and R. Jones, "Condition monitoring and failure
prediction in chaos," IEE Colloquium on Advanced Vibration
The macro-average F1-score for the model was 1.00 and the Measurements, Techniques and Instrumentation for the Early
Prediction of Failure, London, UK, 1992, pp. 3/1-3/9.
weighted F1-score was 1.00. This demonstrates well-rounded
[10] "IEEE Draft Guide for Application of Electric Motors in Class I,
performance across all classes. The recall and the precision Division 2 and Class I, Zone 2 Hazardous (Classified) Locations,"
indicates that the model is highly reliable, and in class- in IEEE P1349/D9.4, June 2011 , vol., no., pp.1-156, 29 June 2011,
imbalanced data too, with the occurrence in the future. doi: 10.1109/IEEESTD.2011.5937008.
[11] M. Tsypkin, "Induction motor condition monitoring: Vibration analysis
technique - A practical implementation," 2011 IEEE International
Electric Machines & Drives Conference (IEMDC), Niagara Falls, ON,
REFERENCES Canada, 2011, pp. 406-411, doi: 10.1109/IEMDC.2011.5994629.
[1] E. Melial "Development and Preliminary Validation of a New Strategy [12] “ISO 10816-3: Evaluation of Machine Vibration by measurement on
to Model the Interaction Between Rotating Machines and Elastic non-rotating parts,” ISO, 1995.
Supporting Structure" paper on 9th IFToMM International Conference
[13] S.P. Mogal, and D.I. Lalwani, "Fault diagnosis of bent shaft in rotor
on Rotor Dynamics J. Clerk Maxwell, A Treatise on Electricity and
bearing system," Journal of Mechanical Science and Technology, vol.
Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.
31, no. 1, pp. 1-3, 2017
[2] ISO 10816 “Mechanical vibration - Evaluation of machine vibration by
measurements on non-rotating parts
[3] T. D. Graybeal, “The nature of vibration in electric machinery,” Trans
actions of the American Institute of Electrical Engineers, vol. 63, no.
10, pp. 712–718, Oct 1944.
[4] C. Li and M. Xu, “Notice of retraction lt;br gt;reliability assessment of
the system subjected to sudden and multiclass gradual failures,” in
2010 2nd International Conference on Computer Engineering and
Technology, vol. 5, April 2010, pp. V5–83–V5–87.
[5] K. Yavanarani, K. Sekar, G. S. S. S. Raj, and S. S. Christabel, “3d
spectral analysis for UGV in analyzing vibration signal,” in
INTERACT 2010, Dec 2010, pp. 138–142.
[6] O. V. Afanaseva, D. A. Pervukhin, A. F. Nyrkov, S. Y. Chernyi, O. K.
Bezyukov, and Y. N. Serditov, “Combustion engines vibro-activity
diagnostic system, using the methods of similarity theory and

Page 8 of 8 - Integrity Submission Submission ID trn:oid:::30744:108023393