1 of 45

���FAULT DIAGNOSIS OF BROACHING MACHINE USING MACHINE LEARNING APPROACH�

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

1

RMD SINHGAD SCHOOL OF ENGINEERING WARJE-58

DEPARTMENT OF MECHANICAL ENGINEERING

Group No: 27

​

Guided By:

Prof. N. C. Dhobale

Presented by

  1. Vanve Vishvtej Vishnu 71802725G
  2. Waghule Rushikesh Dnyandeo 71802738J
  3. Yadav Prashant Harishchandra 71802752D
  4. Jadhav Sudarshan Tatyasaheb 71802206J

2 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

2

Figure No.

Figure Description

Figure 1

The Broaching operation

Figure 2

2.1 Work shapes that can be cut by external broaching

2.2 Work shapes that can be cut by internal broaching

Figure 3

Worn cutting tool showing the principal locations and types of wear that occur.

Figure 4

  1. Crater wear on a cemented carbide tool
  2. Flank wear on a cemented carbide tool

Figure 5

The broach : a) Terminology of the tooth geometry

b) A typical broach used for internal broaching.

Figure 6

Block Diagram of FFT Analyzer

Figure 7

FFT Analyzer

Figure 8

  1. Data acquisition system
  2. Interfacing block diagram of adxl-335 accelerometer with Arduino

Figure 9

Flow chart for steps of Machine Learning Module

Figure 10

Signal acquisition in time domain for all machining operations

List of Figures

3 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

3

Figure No.

Figure Description

Figure 11

Data reading

Figure 12

Time Domain Graphs

Figure 13

Result of decision tree for feature selection

Figure 14

Confusion Matrix for Random Forest Classifier

Figure 15

Line Plot of Flank Wear Condition

Figure 16

Line Plot of Notch Wear Condition

Figure 17

Line Plot of Crater Wear Condition

Figure 18

Line Plot of Abrasive Wear Condition

4 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

4

NOMENCLATURE

Symbol

Abbreviation(Full Form)

FFT

Fast Fourier Transform

DAQ

Data Acquisition System

MEMS

Micro Electro Mechanical system

A

Amplitude of Vibration

f

Frequency of Vibration

RPM

Rotation per minute

ANN

Artificial Neural Network

5 of 45

Contents

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

5

Contents

Slide No.

Introduction

​

Types of Broaching

​

Need of fault Diagnosis

​

Literature Review

​

Problem Statement

​

Objective

​

Methodology

​

Design of DAQ

​

Proposed work & Plan

​

Experimentation

​

Result

​

Conclusion

​

Scope of future work

​

References

​

6 of 45

INTRODUCTION

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

6

  • Broaching is performed using a multiple-teeth cutting tool by moving the tool linearly relative to the work in the direction of the tool axis
  • The machine tool is called a broaching machine, and the cutting tool is called a broach
  • Broaching is a highly productive method of machining
  • Advantages include good surface finish, close tolerances, and a variety of work shapes.
  • There are two principal types of broaching: External (also called surface broaching) and internal

7 of 45

Types of Broaching

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

7

  • External broaching is performed on the outside surface of the work to create a certain cross-sectional shape on the surface
  • Internal broaching is accomplished on the internal surface of a hole in the part
  • Accordingly, a starting hole must be present in the part to insert the broach at the beginning of the broaching stroke
  • The basic function of a broaching machine is to provide a precise linear motion of the tool past a stationary work position
  • Most broaching machines can be classified as either vertical or horizontal machines

8 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

8

  • The vertical broaching machine is designed to move the broach along a vertical path
  • The horizontal broaching machine has a horizontal tool trajectory
  • Most broaching machines pull the broach past the work
  • One exception is a relatively simple type called a broaching press, used only for internal broaching, that pushes the tool through the work-part
  • Another exception is the continuous broaching machine, in

which the work-parts are fixtured to an endless belt loop and moved past a stationary broach.

9 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

9

Fig 1 The Broaching operation

Cross-hatching indicates the surfaces broached.

Fig 2.1 Work shapes that can be cut by external broaching

Fig 2.2 Work shapes that can be cut by internal broaching

10 of 45

Types of wear in a cutting tool

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

10

​

  • Gradual wear occurs at two principal locations on a cutting tool: the top rake face and the flank
  • Accordingly, two main types of tool wear can be distinguished: crater wear and flank wear
  • Crater wear consists of a cavity in the rake face of the tool that forms and grows from the action of the chip sliding against the surface
  • High stresses and temperatures characterize the tool–chip contact interface, contributing to the wearing action
  • The crater can be measured either by its depth or its area

11 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

11

  • Flank wear occurs on the flank, or relief face, of the tool.
  • It results from rubbing between the newly generated work surface and the flank face adjacent to the cutting edge
  • Flank wear is measured by the width of the wear band, This wear band is sometimes called the flank wear land
  • An extreme condition of flank wear often appears on the cutting edge at the location corresponding to the original surface of the workpart. This is called notch wear
  • Notch wear occurs because the original work surface is harder and more abrasive than the internal material, which could be caused by work hardening from cold drawing or previous machining

​

12 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

12

  • Mechanisms that cause wear at the tool–chip and tool–work interfaces in machining :
  • Abrasion, Adhesion,
  • Diffusion, Chemical reactions,
  • Plastic deformation

Fig 3 : Worn cutting tool showing the principal locations and types of wear that occur.

13 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

13

Fig 4 a) Crater wear , b) Flank wear on a cemented carbide tool, as seen through a toolmaker’s microscope

14 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

14

Fig 5 The broach 5a) Terminology of the tooth geometry

5b) A typical broach used for internal broaching.

15 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

15

  • FFT Spectrum Analyzer

The instrument which converts the input signal, with time as an independent variable, into frequency spectrum and displays it in graphical form is called as spectrum analyzer or FFT analyzer

Fig 6 BLOCK DIAGRAM OF FFT Analyzer

16 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

16

We have used Computer based FFT Analyzer as shown below

Fig 7 FFT Analyzer

17 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

17

​

  • Tool wear/breakage lead to
    • Deterioration of machining accuracy.
    • Poor quality and variation in dimension of the product
  • Need of fault diagnosis of broaching machine
    • To provide a health condition monitoring
    • Continuous monitoring of certain parameters
    • Promises highly productivity
    • Reduced maintenance cost
    • Saving idle time
    • Enhancing tool and machine life
  • Need of fault diagnosis

18 of 45

LITERATURE REVIEW

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

18

​

​

19 of 45

LITERATURE REVIEW

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

19

​

​

20 of 45

LITERATURE REVIEW

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

20

​

​

21 of 45

PROBLEM STATEMENT

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

21

  • FAULT DIAGNOSIS OF BROACHING MACHINE USING MACHINE LEARNING APPROACH

22 of 45

OBJECTIVE

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

22

​

​

​

​

To develop DAQ for vibration acquisition using MEMS accelerometer and open source hardware

To correlate vibration and broach condition for different machining parameters.

To classify broach faults using machine learning approach

23 of 45

METHODOLOGY

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

23

​

​

Broaching machine

Accelerometer & FFT analyzer

Data acquisition system

Sampled signal

Machine learning approch

Feature extraction

Feature selection

Feature classification

Fault diagnosis result

24 of 45

In-house design of DAQ�

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

24

​

​

​

​

​

Fig 8a Module of DAQ

Fig 8b Interfacing diagram of

ADXL335 with Arduino uno

25 of 45

MACHINE LEARNING

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

25

  • Machine learning is a branch of artificial intelligence where the machine learns the input given by the user and then analyses the features from the number of inputs given.
  • It is a scientific study of algorithms and statistical models that computer systems use to perform different tasks.
  • Machine learning is related to computational statistics which focus on making predictions using computers.

​

​

26 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

26

Steps In Machine Learning Module

Fig 9 Flow Chart

27 of 45

Proposed Work & Plan

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

27

  • To identify different modes of broach failures.
  • . To design experiment for vibration data acquisition
    • Decide workpiece material
    • Decide machining parameters as speed, feed, depth of cut
    • Decide faults to be considered for experimentation
  • To carry out machining and vibration acquisition using in house built DAQ
  • To plot time domain graph and extract statistical features

(14 descriptive statistics)

  • To analyze statistical features using Machine Learning approach aiming for fault classification

28 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

28

Fig 10 Experimental setup at Industrial Broaching Machine site

29 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

29

Fig 11 Data reading

30 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

30

  • The vibration signal was captured for all 3 operations and respective time domain plots acquired in Arduino software .
  • It is observed in figure that the acceleration amplitude varies and rises as the fault condition changes

31 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

31

Results for Testing : Time domain graphs

  • Signal acquisition in time domain for all machining parameters

Fig 12 Time Domain Graphs

32 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

32

Fig 13 Result of decision tree for feature selection

33 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

33

Nomenclature:

  • MN: Mean, SE: Standard Error, MDN: Median, SD: Standard Deviation,
  • KT: Kurtosis, SK: Skewness, RG: Range, MIN: Minimum, MAX: Maximum, A, B, C are the tool conditions as already.
  • The feature selection is the most significant step in machine learning approach as it selects most significant features amongst all.
  • It reduces dimensionality; speed up the classification algorithm; increases classification accuracy and makes the results more comprehensible

34 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

34

Sr No.

Constraints used for classification

( 9 features selected by decision tree

Algorithm Type

Decision Tree Classifier

Gradient Boosting Classifier

​

Random Forest Classifier

1.

MN+SE+MDN+

SD+KT+SK+

Range + MIN +

MAX Condition

Accuracy

91%

93.3%

93.3%

RESULT TABLE

35 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

35

Fig 14 Confusion Matrix for Random Forest Classifier

36 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

36

Fig 15 Line Plot of Flank Wear Condition

37 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

37

Fig 16 Line Plot of Notch Wear Condition

38 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

38

Fig 17 Line Plot of Crater Wear Condition

39 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

39

Fig 18 Line Plot of Abrasive Wear Condition

40 of 45

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

40

COST ESTIMATION

41 of 45

Conclusion

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

41

  • The development data acquisition system that uses an open source prototyping platform (Arduino) which significantly reduce implementation cost while keeping data’s integrity.
  • The proposed system will be validated by comparing its results commercial wired system through an experimental in laboratory
  • It is observed that low cost accelerometer has a good performance ,especially for low Signal - Noise ratios

i.e high acceleration values

  • In frequency domain analysis ,the result should be identical to those obtained by commercial accelerometer

42 of 45

Scope of future work

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

42

  • Works has been carried out using conventional method however only few experimental analysis has been performed for broaching machine.
  • Application of machine learning technique for cutting tool fault diagnosis.
  • Broach Cutting tool with faulty condition to be consider and these faults classify with machine learning technique.
  • Effective combination of feature classification in fault diagnosis for cutting tool.
  • Real time implementation of Machine learning with hardware and software.

43 of 45

References

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

43

  • Michelle Thompson, Mark McCormick, ‘Monitoring machine health’, midstreambusiness.com, January 2013
  • Jonathan C. Young, ‘Vibration Analysis using a MEMS Accelerometer’, Thesis submitted for the degree of MS in Electrical Engg,Naval, Postgraduate School, Monterey, California, December 2006
  • Arturo Garcia-Perez, Konstantinos N. Gyakis, ‘Vibration Analysis as a Diagnosis Tool for Online Health Monitoring of Industrial Machines’, International Journal of shock and vibrations, Procedia Engineering 122 ( 2015 ) 215 – 195.
  • Mark Looney, ‘An Introduction to MEMS Vibration Monitoring’, Analog Dialogue 48-06, June (2014).

44 of 45

References

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

44

  • K. N. Ede, E. A. Ogbonnaya, M. T. Lilly, S. O. T. Ogaji, S. D. probert, ‘Vibration Monitoring of Rotating Systems’, Engineering, 2010, 2, 46-54.
  • Subimal Bikash Chaudhury, Mainak Sengupta, Kaushik Mukherjee, Vibration Monitoring of Rotating Machines Using MEMS Accelerometer, International Journal of Scientific Engineering and Research (IJSER), ISSN (Online): 2347-3878, Volume 2 Issue 9, September 2014.
  • Namdev A. Patil, Laukik P. Raut, ‘Vibration Analysis of CI Engine using FFT Analyzer’, International Journal of Research in Engineering and Technology, 2016

45 of 45

THANK YOU

DEPARTMENT OF MECHANICAL ENGINEERING, RMDSSOE BE (2021)

45