A Minor Project Report

On

(BEARING VIBRATIONAL FAULT DIAGNOSIS USING MACHINE

LEARNING)

Submitted by
SARANSH BAIRAGI
(0901AU201028)

Under the guidance of

Prof. Vaibhav Shivhare
(Assistant Professor)

Department of Mechanical Engineering,

MITS, Gwalior

DEPARTMENT OF MECHANICAL ENGINEERING

MADHAV INSTITUTE OF TECHNOLOGY & SCIENCE
GWALIOR-474005 (M.P)
April, 2023
Madhav Institute of Technology & Science, Gwalior (M.P.)
(A Govt. Aided UGC Autonomous Institute & NAAC Accredited, Estd. in 1957, Affilated to RGPV Bhopal)

STUDENT DECLARATION
I hereby certify that the Minor Project entitled Bearing Vibrational Fault Diagnosis Using
Machine Learning which is being submitted in the Department of Mechanical Engineering is
a record of my own work carried out under the supervision and guidance of Prof. Vaibhav Shivhare
Assistant Professor, Department of Mechanical Engineering, Madhav Institute of Technology &
Science, Gwalior.

All information in this document has been obtained and presented in accordance with academic
rules and ethical conduct. I have fully cited and referenced all material and results that are not
original to this work.

To the best of my knowledge the material presented in this project has not been submitted to any
other place (i.e. institute, university, organization) as thesis, report.

Date: 28/04/2023
Place: Gwalior SARANSH BAIRAGI

This is to certify that the above statement made by the candidate is correct to the best of my knowledge
and belief.

Guided by

Prof. Vaibhav Shivhare Dr.Manoj Kumar Gaur

Assistant Professor H.O.D
Department of Mechanical Engineering Mechanical Department
MITS, Gwalior MITS, Gwalior

i
 ACKNOWLEDGEMENT
I would like to express my sincere appreciation to my supervisor Prof.Vaibhav Shivhare
Mechanical Departmental H.O.D Dr.Manoj Kumar Gaur and Dr.R.K Pandit for his guidance,
encouragement, and support throughout the course of this project work. It was an invaluable
learning experience for me to be one of their students. From them I have gained not only extensive
knowledge, but also a careful research attitude.

I am also thankful to my Batchmates for his cooperation with me in facilitating the

infrastructure and lab facility during my Minor Project work. I am also thankful to all my friends,
which have contributed and helped me directly or indirectly in bringing this project to a successful
completion. Finally, I am highly indebted to my parents, my family members for extending every
possible support with great inspiration and numerous blessing towards me.

SARANSH BAIRAGI
0901AU201028

ii
 ABSTRACT

The purpose of the Minor-Project was to analyze Infrared thermography and Vibrations of bearing
for detecting faults in machine systems. However, sometimes its performance is reduced due to
the peculiar noise generated by the machine and string noise. To overcome this, we will take
advantage of 2D-DWT; A Fourier transform based signal processing tool that reduces noise and
provides better results in thermography. The internet is so advanced today that all information is
at your fingertips. Machine component maintenance, also called condition monitoring, is currently
performed by integrating electronic computing power and compatible measurement hardware.
Software such as LabVIEW and Python bridge the gap between the computer and the measurement
hardware. Remote operation and maintenance of heavy machinery is possible in the near future
and thus competes with various factors such as unavailability of people, environmental factors
Maintenance of machinery is a very essential work that every industry needs to do regularly.
Operation, control and maintenance of heavy equipment by one person is an impossible task.
Regular servicing is essential for the successful and trouble-free operation of machines. Condition
monitoring is one such machine maintenance method where a fault on the machine is diagnosed
using IRT.

Data Science helps to utilise data to construct algorithms and create programs to solve any
problems. For creating programs, a programming language python can be used. Python will help
to create problems and algorithms which will be further analyzed by Machine learning. Artificial
intelligence will be further used to improve the data which was previously analysed by machine
learning and by this, the experimental data can be used to give a desired and the fault detection
results.

iii
 TABLE OF CONTENTS

Title Page No.

CANDIDATE’S DECLARATION i

ACKNOWLEDGEMENT ii

ABSTRACT iii

TABLE OF CONTENTS iv

1. CHAPTER 1 : Introduction v-vii

1. History of ball bearing v

2. Present Day vi
3. Aim & Objectives vii

2. CHAPTER 2 : Literature Review/Background study viii

3. CHAPTER 3: Methodology/ Materials and Methods ix

4. CHAPTER 4: Results and Discussion/Model/Prototype and its working x

5. CHAPTER 5: Conclusion xi

6. CHAPTER 6: Future Scope/Future work xii

Reference xiii

TURNITIN certificate of the report xiv

iv
CHAPTER 1: INTRODUCTION

History of ball bearing:

The first and the most important device which was based on a ball-bearing element was discovered in the Roman
Empire about 40AD. In that device, rotating table and the balls were present and rotating table. By using that
table, people eating at that table simply turn the table to serve the meal easily. The next design for a ball bearing
used in a rotary device or machine was found after 1500 years was discovered by "Leonardo".
After that , almost 100 years ago Galileo also mentioned a ball bearing design. Until 1791 there was no at all any
patent field and design for a ball bearing and then the 1st patent on ball bearing was sent to "Philip Vaughan" in
1971. In ancient times when the wheel was not discovered at the start of human civilization, the concept of bearing
had already been used. A rolling vehicle which was used to carry high load by dragging the vehicle. In that vehicle
the rolling logs were placed beneath heavy objects which makes the dragging of high loads easier. Different sort
of liquids used as a lubricants for smooth functioning.

Types of Bearing:
Ball Bearings: Ball bearings have balls as rolling elements.

Deep-Groove Ball Bearings: This is the type of ball bearing with less torque

Angular Contact Ball Bearings: In this type of ball bearing things are on certain angle and force is transferred
in that

Angle Roller Bearings: Roller bearings are of point contact and have circular cross-section

Spherical Roller Bearings: Spherical Roller Bearings have spherical cross-section and are of Roller Bearing
types.

Cylindrical Roller Bearings: These use linear point of contact between surface.

v
Present Day:

Various machine learning algorithms are introduced nowadays and researchers are still working on the various
methods of AI and ML techniques for detecting fault.

In 1869 "Jules Suriray" designed the ball bearing and that the bearing were used in bicycles. This improved design
and development of new different varieties of metal ball bearings. "Sven Winquist" gave a new bearing design
which was a self-aligning design. During the 20th century, the research continued to change and improve the
bearings. The requirement for all mechanical systems is bearings which could support forces and combined loads
applied axially and radially in the system and reduces a large amount of friction to get an efficient machine system.
In a ball bearing, balls are the rolling element which is used to separate the races of bearing.

The main function of a ball bearing is the reduction in friction in rotational direction and to provide support at the
loads. As the failure in ball bearing can lead to destroy Machine, the condition monitoring of the system is very
much important which is used for identify the fault in bearing in the system.

Aim And Objective:

For detecting and identifying the defects in bearing, vibrational analysis is used. Monitoring the health of the
machine each and every time is essential which can be done by condition based maintenance or predictive
maintenance. Determining natural frequency damping and shapes of structure, experiment model analysis is used.
For reducing the effects of failure and identify feasible improvements risk priority number is analysed in every
machine before manufacture. The Caused may be due to improper lubrication, faulty installation or overload that
can be due to misalignment, shaft detection, or tilted inner race. Artificial intelligence algorithms can empower
manufacturing industry for the health of bearing.

For detecting the fault diagnosis by processing the data collected by the sensor and combining deep learning or
machine learning. Sensors are installed at different positions to collect vibration signals of the bearing. Vibrational
analysis consist of the fault diagnosis,machine learning in CBM, vibration and shock isolation, time and frequency
analysis, vibration and monitoring machinery and the most important tool and condition based monitoring. After
studying and getting knowledge of vibrational analysis the summarization of basic and beneficial methods and
principles of the machine learning based rolling bearing fault diagnosis takes place which helps to analyse
experimental results. After performing the experiments, experimental data is analysed and the knowledge of data
is extracted with the help of machine learning.

vi
 CHAPTER 2 : Literature Review/Background study

We went through various Machine Learning process and algorithms followed by different types of Machine
learning through which We can apply the algorithms for finding out the fault in the bearing in any machinery
component with the help of experimental dataset.

Also went through vibrational analysis by watching a video lecture series of “Machinery fault diagnostic and
signal processing” which was delivered by professor A.R. Mohanty from IIT Kharagpur. Saw the lectures on a
Python based lectures on prognostic and health management using vibrational signal which includes signal
processing future extraction and feature selection. Analysed the research paper of topics support vector machine
commonly known as SVM, KNN algorithm, Various method of Unsupervised learning in machine condition
monitoring and fault diagnosis, artificial intelligence for fault diagnosis of rotating machinery, investigated the
techniques for rolling bearing, fault diagnosis using machine learning algorithms, A new statistical features based
approach for bearing fault diagnosis using vibration signals.

Bearing faults are one of the major causes of machinery breakdowns and failures. Early detection and diagnosis
of bearing faults are crucial for preventing catastrophic failures and minimizing downtime. Vibration analysis has
been widely used as a non-invasive and effective method for bearing fault detection. In recent years, machine
learning techniques have emerged as promising tools to enhance the accuracy and efficiency of bearing fault
analysis. This literature review provides an overview of the research conducted on bearing fault analysis using
vibrations and machine learning. The review highlights the key methodologies, challenges, and advancements in
this field.

Features and methods used for fault detection Frequency-domain analysis: Fourier analysis, spectral analysis, and
envelope analysis Wavelet transform-based techniques: Continuous wavelet transform, discrete wavelet
transform, and their applications Statistical techniques: Kurtosis, skewness, and other statistical measures
Machine Learning Techniques for Bearing Fault Analysis Supervised learning algorithms: Support Vector
Machines (SVM), Artificial Neural Networks (ANN), Random Forests (RF) Unsupervised learning algorithms:
Clustering techniques, such as K-means and hierarchical clustering Deep learning algorithms: Convolutional
Neural Networks (CNN), Long Short-Term Memory (LSTM) networks Feature selection and extraction methods
in machine learning for bearing fault analysis Integration of Vibration Analysis and Machine Learning Feature
extraction from vibration signals for machine learning-based fault detection Ensemble methods combining
multiple classifiers for improved fault detection performance Fault classification and identification using machine
learning algorithms

Real-time monitoring and online fault detection using machine learning techniques Challenges and Future
Directions Data acquisition and preprocessing challenges Selection of appropriate features for fault detection
Generalization and robustness of machine learning models Integration of domain knowledge with machine
learning approaches Utilization of emerging technologies, such as Internet of Things (IoT) and edge computing
Conclusion Summary of key findings and contributions Potential applications and benefits of bearing fault
analysis using vibrations and machine learning Future research directions and recommendations References List
of cited studies, research papers, and relevant literature Note: This two-page literature review provides a broad
outline of the topics to be covered. However, for a comprehensive literature review, it is important to review and
analyze multiple relevant studies, cite specific examples, and provide a critical evaluation of the findings.

vii
CHAPTER 3: Methodology/ Materials and Methods

The first step taken to find the fault detection in bearing is to learn about the bearing, the parts of
bearing, the defects in bearing, the types of bearing and have the deep knowledge about bearing.
The next step taken is to learn about vibrational analysis and machinery fault diagnosis and signal
processing which includes the maintenance principles, failure mode affect and criticality analysis,
fault diagnostic and prognostic, machine learning for CBM, basics of machinery vibration,
vibration and shock isolation, rotar dynamics, time and frequency domain analysis, signal
Modulation and Beats, Computer aided data acquisition, Vibration monitoring, noise monitoring,
rotational speed, measurements, errors in measurements, signal condition and filtering,
introduction to fault in rotating machines, balancing, miss alignment , Journal and anti friction
bearings, machinery diagnostic chart, fault in electrical machine, oil analysis, tool condition
monitoring, experimental model analysis, introduction to failure analysis, Railway locomotive,
noise and vibration monitor and the future of condition based maintenance.

After having the deep knowledge about vibrational analysis and bearing we can perform the
experiments in Laboratories and the data of the experiment can be used in creating programs with
the help of python programming language which can be further used to utilise data to construct
all algorithms with the help of data science and it can be further analysed by machine learning.
Artificial intelligence is used to improve the data which is analysed by machine learning and by
this a fault can be detected by using the experiment data with the help of artificial intelligence.

CHAPTER 4: Results and Discussion/Model/Prototype and its

viii
working.

In Machine leaning algorithms, the 90% of the dataset is carried out for the training and the rest 10% is carried
out to test the dataset by Confusion matrix of classification problem or by any other problem.

This methodology section outlines the steps involved in conducting bearing fault analysis using vibrations and
machine learning techniques. The methodology encompasses data acquisition, preprocessing, feature extraction,
machine learning model development, and evaluation. By following this methodology, researchers can effectively
analyze bearing faults and enhance the accuracy and efficiency of fault detection.

1. Data Acquisition:
• Identify the target machine and bearing to be monitored.
• Install appropriate sensors to capture vibration signals.
• Determine the sampling frequency and duration of data acquisition.
• Conduct experiments or collect real-time data from the machine under normal and faulty
conditions.
• Ensure the data acquisition setup is properly calibrated and free from any external interferences.

2. Preprocessing:
• Remove any noise or unwanted artifacts from the acquired vibration signals.
• Perform signal filtering techniques such as low-pass, high-pass, or band-pass filtering to eliminate
irrelevant frequencies.
• Correct for baseline shifts and remove any trends or biases in the data.
• Normalize or scale the vibration signals to ensure consistent data representation across different
sensors or experiments.
• Split the data into training and testing sets for machine learning model development and
evaluation.

3. Feature Extraction:
• Extract relevant features from the preprocessed vibration signals.
• Time-domain features: Compute statistical measures like mean, standard deviation, kurtosis,
skewness, etc.
• Frequency-domain features: Apply Fourier Transform or other spectral analysis techniques to
obtain frequency-related features such as peak amplitudes, dominant frequencies, and spectral
centroid.
• Wavelet transform-based features: Perform continuous wavelet transform or discrete wavelet
transform to extract time-frequency representations of the vibration signals.
• Select appropriate feature extraction techniques based on the specific characteristics of the bearing
fault being analyzed.

4. Machine Learning Model Development:

• Select suitable machine learning algorithms based on the problem type (classification, regression)
and available data.
• Divide the training data further into training and validation sets for model training and tuning.
• Train the selected machine learning models using the training data and optimize model parameters.
• Evaluate different algorithms and variations to identify the best-performing model.

ix
 • Consider ensemble methods or combination of multiple models for improved accuracy and
robustness.

5. Model Evaluation:
• Use the testing dataset to evaluate the performance of the trained machine learning models.
• Calculate performance metrics such as accuracy, precision, recall, F1-score, and area under the
curve (AUC).
• Analyze the confusion matrix to assess the model's ability to correctly classify fault types.
• Compare the results with baseline methods or existing techniques to demonstrate the effectiveness
of the proposed methodology.
• Conduct cross-validation or other validation techniques to ensure the model's generalization and
reliability.

6. Iterative Refinement:
• Analyze the shortcomings and limitations of the developed methodology.
• Identify potential sources of error or bias in the analysis and propose appropriate mitigations.
• Fine-tune the preprocessing steps, feature extraction techniques, or machine learning models based
on the initial evaluation results.
• Repeat the steps of data acquisition, preprocessing, feature extraction, model development, and
evaluation to refine the methodology. Note: This methodology provides a general framework for
conducting bearing fault analysis using vibrations and machine learning. Researchers should adapt
and customize the methodology based on the specific requirements of their study and the available
resources.

x
CHAPTER 5: Conclusion

A machine learning based fault diagnosis method for ball Bearings based on time domain data and frequency
spectrum Data is found successfully. This work can be done only with a specific speed range. Temperature data
can also be used for Developing a better condition monitoring system. With the Use of additional data, the
condition of the system can be Predicted more accurately. With the help of Confusion matrix, the test data is
verified for the respective dataset and methods.

Using various methods and principles many more dataset can also be checked for diagnosing the fault in any
machinery component. Data pre processing plays a crucial role for the same mainly for removing the unwanted
columns and removing missing values.

CHAPTER 6: Future Scope/Future work

xi
With the help of Machine learning, Deep learning and Artificial Intelligence, many more problems can also be
resolved with the help of various methods and algorithms.
Not only bearings but also gears and other machinery components fault can also be diagnosed very easily.
The hardware implementation of the proposed system for real-time condition monitoring and the extension in this
system for monitoring the different types of faults of different intensities and signals will be the scope in the
future work.

REFERENCES

xii
[1] Toliyat, H.A.; Nandi, S.; Choi, S.; Meshgin-Kelk, H. Electric Machines: Modeling, Condition Monitoring,
and Fault Diagnosis; CRC Press: Boca Raton, FL, USA, 2012.
[2] "Double- Row Angular Contact Ball Bearings". Archived from the original on 11 May 2013.
[3] Louis Baudry de Saunier, Histoire générale de la vélocipédie [General history of cycling] (Paris, France:
Paul Ollendorff, 1891), pages 62–63
[4] leonardodavincisinventions.com/mechanical-inventions /leonardo-da-vinciball-bearing
[5] Tamilselvan P, Wang P. Failure diagnosis using deep belief learning based health state classification.
Reliab Eng Syst Saf 2013;115:124–35. DOI: 10.1016/j.ress.2013.02.022.
[6] Janssens O, Slavkovikj V, Vervisch B, Stockman K, Loccufier M, Verstockt S, et al. Convolutional Neural
Network Based Fault Detection for Rotating Machinery. J Sound Vib 2016; 377:331–45. DOI:
10.1016/j.jsv.2016.05.027
[7] Shen Zhang, Shibo Zhang, Bingnan Wang, and Thomas G. Habetler, Machine Learning and Deep
Learning Algorithms for Bearing Fault Diagnostics – A Comprehensive Review.
[8] Bo Li, Student Member, IEEE, Mo-Yuen Chow, Senior Member, IEEE, Yodyium Tipsuwan, Student
Member, IEEE, and James C. Hung, Fellow, IEEE, Neural-Network-Based Motor Rolling Bearing Fault
Diagnosis.
[9] Mukesh Bhadane, K.I. Ramachandran, “Bearing Fault Identification and Classification with
Convolutional Neural Network” 978-1- 5090-4967- 7/17/$31.00 © 2017 IEEE
[10] A python based tutorial on prognostics and health management using vibration signal: signal
processing, feature extraction and feature selection and also a Machine learning Lecture series.
[11] OpenAI. 2022. Chatgpt. https://chat.openai. com.
[12] Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin,
Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to
follow instructions with human feedback. Advances in Neural Information Processing Systems,
35:27730–27744.
[13] Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix,
Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation
language models. arXiv preprint arXiv:2302.13971. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra.
[14] Alisa Liu, Noah A Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. Self-instruct: Aligning
language model with self generated instructions. arXiv preprint arXiv:2212.10560. Jason Wei.
[15] Maarten Bosma, Vincent Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and
Quoc V Le. 2022. Finetuned language models are zero-shot learners. In International Conference on Learning
Representations. Honglin Xiong, Sheng Wang, Yitao Zhu, Zihao Zhao, Yuxiao Liu, Qian Wang, and Dinggang
Shen.
[16] Doctorglm: Fine-tuning your chinese doctor is not a herculean task. Aohan Zeng, Xiao Liu, Zhengxiao Du,
Zihan Wang, Hanyu Lai, Ming Ding, Zhuoyi Yang, Yifan Xu, Wendi Zheng, Xiao Xia, Weng Lam Tam, Zixuan
Ma, Yufei Xue, Jidong Zhai, Wenguang Chen, Zhiyuan Liu, Peng Zhang, Yuxiao Dong, and Jie Tang. 2023.
[17] GLM-130b: An open bilingual pretrained model. In The Eleventh International Conference on Learning
Representations (ICLR).
[18] Zhihong Chen, Junying Chen, Hongbo Zhang, Feng Jiang, Guiming Chen, Fei Yu, Tiannan Wang, Juhao
Liang, Chen Zhang, Zhiyi Zhang, Jianquan Li, Xiang Wan, Haizhou Li, and Benyou Wang. 2023. Llm zoo:
democratizing chatgpt. https://github. com/FreedomIntelligence/LLMZoo.
[19] Aidan Gilson, Conrad W Safranek, Thomas Huang, Vimig Socrates, Ling Chi, Richard Andrew Taylor,
David Chartash, et al. 2023. How does chatgpt perform on the united states medical licensing examination? the
implications of large language models for medical education and knowledge assessment. JMIR Medical Education,
9(1):e45312.
[20] Yunxiang Li, Zihan Li, Kai Zhang, Ruilong Dan, and You Zhang. 2023. Chatdoctor: A medical chat model
fine-tuned on llama model using medical domain knowledge.

13 | P a g e
14 | P a g e
15 | P a g e