11 IV April 2023

https://doi.org/10.22214/ijraset.2023.50279
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue IV Apr 2023- Available at www.ijraset.com

Gear Defect Detection using Machine Learning

Prof. Archana Renushe1, Rutuja Kesare2, Rohini Kumbhar3, Pooja Kumbhar4
1, 2, 3, 4
Dr. Daulatrao Aher College of Engineering, India

Abstract: This proposed work introduces the application that is used to perceive the condition of system equipment. In machines,
equipment is a very important component on every occasion it is damaged then it immediately affects on gadget's reliability. The
objective of this project is to increase a system that shows the type of tools or condition of tools used to locate whether or not it's
miles faulty or non-faulty manner. Convolution neural network technique might be used to locate the situation of tools and This
approach is accomplished via picture processing. It takes the photograph manually from datasets or via the live digicam. The
actual result of this Proposed work is to show the pick-out gear picture in which the form of gear is defective or non-defective.
We explored and in comparison various CNN networks for item detection using real facts furnished by the consumer. The
surface illness detection and counting the variety of equipment teeth manually isn't always correct so to clear up the trouble.
Keywords: defect detection, image processing

I. INTRODUCTION
The reliability of the machine is directly impacted by the failure of the system's primary component, the gear, which is the
fundamental part of mechanical transmission systems. Thus, gear inspection is a crucial responsibility in machine work. A gear train
is a sequence of two or more pairs of gears working together. As it is crucial to find gear faults, this can be done in a variety of
methods, including by examining the frequency domain of a condition known as torque in ratio. Gear fault identification has
limitations in that it cannot recognize all sorts and categories of gear faults, even using time-frequency, vibration, and noise produced
by the faulty. This research creates a gear defect-detecting application device that uses CNN image processing to address the
aforementioned issues. For this image processing, we first choose an image from the data set, compare it to a healthy gear image
using the CNN method, and then determine if the result is faulty or not. Because it lowers maintenance costs, it is particularly
beneficial for railroad equipment and other devices. Additionally, it aids in resolving the gear defect issue that compromises the
reliability and safety of railroad operations. It plays a crucial role in the industry in determining whether or not the gear is defective so
that material production can continue gear.

II. LITERATURE SURVEY

1) Gear Assembly Fault Analysis Using Dual-Tree Complex Wavelet Transform Kashwan, K.R.,Senior Member. Deepika S [2016]
The acoustic signal produced by malfunctioning gears in a gear mesh is typically loud and non-stationary. The broken gears could
multiply and worsen the damage to the whole gear mechanism. This work describes a dual-tree complex wavelet transform-based
method for diagnosing gear faults. The reference signal is the acoustic signal from the sound gear mesh. The analysis is done on gear
that has one or more teeth with ingrained flaws. It is possible to accurately estimate the angles between two or more damaged teeth.
By running simulations, the suggested method is utilized to assess the efficacy of detection. Measurements are made of the parameters
mean, standard deviation, auto correlation, dynamic range, and crest factor. Gear fault identification and defect severity estimation
employ acoustic signals. It assists in avoiding gear tooth fractures and motors gear health.

2) A Bevel Gear Quality Inspection System Based on Multi-Camera Vision Technology.Dexing Zhong2 and Jiuqiang Han4, Ruiling
Liu1,Hongqiang Lyu3[2012]
Automotive bevel gear surface defect detection and dimension measurement by guide inspections are expensive, ineffective, low
velocity, and occasionally accurate. A synthetic bevel gear excellent inspection machine built on multi-digital camera vision
technology is advanced to address such problems. The tool may concurrently measure gear dimensions and find surface flaws.
Neighbourhood Average Difference (NAD), Circle Approximation Method (CAM), and Fast Rotation-Position (FRP) are three green
methods that are put forth. The tool can identify damage such as knocks, dents, scratches, cracks, porosity, and repeated spline-slicing.
The precision of size is about 40-50 m, and the smallest noticeable disturbance is 0.4 mm. Only 1.3 seconds are needed for one
inspection method. Precision and speed are sufficient for real-time online inspection in the production of bevel equipment.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2906
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue IV Apr 2023- Available at www.ijraset.com

3) Experiment research on the Metal Magnetic Memory in gear micro-crack detection. Lingli Cui2, Jianyu Zhang3, Lixin Gao4, and
Yonggang Xu5,Chenhui Kang1[2015]
A form of the magnetic flux leakage signal of the equipment's early local micro-fracture is generated from the magnetic flux leakage
detection theory of magnetic dipole. The characteristics of the crack area, such as its height within the tangential issue of the magnetic
flux leakage sign, its peak within the gradient, and its bypass 0 points in the daily element, are examined. The evaluation findings
provide the Metal Magnetic Memory utility's fundamental theoretical underpinnings. Finally, utilizing both static testing and dynamic
detection under load, the detection of micro cracks at the side of genuine equipment is implemented entirely based on the notion
presented above. Analyzed are the detection role and load effects on detection outcomes. The efficiency of the peak inside the
gradient of commonplace magnetic flux leakage sign normal element in the tools magnetic flux leakage signal version is
demonstrated by the detection results, and the viability of the magnetic reminiscence testing on equipment fault is also demonstrated.

4) A Defect Status Detecting Method for External Gear in Railway.Fuqiang Li2,Yanfeng Li3,Xuyang Cao5,Chuangang Wang1,
Houjin Chen4[2014]
Railway gear is a crucial component. The gear's reputation for disrepair affects the safety and quality of train travel. An automatic and
quantitative way to determine the disordered state of external tools is suggested in this work. For the segmentation of the meshing
position in the equipment teeth, a -degree method is first presented. Then, to find the floor flaws, adaptive threshold, and formative
evaluation are merged. The suggested method is tested on 140 equipment enamel photographs. The meshing area's proximity overlap
is zero.87. The strategy for detecting illnesses performs better overall than certain related techniques.

III. PROPOSED WORK

In the paper, a system is proposed for finding gear defect detection. In which we provide two data sets train datasets and test datasets.
train datasets are trained by applying a convolution neural network and Gear image and providing labels to each image. Gear images
from test datasets are used as input images going to various classifications like convolution, pooling, fully connected, and soft-
max.TensorFlow is also used in this process to make, it easy to load image datasets and create a CNN fitting the data. we also find out
the type of gear and calculate the teeth of the gear .but the drawback of the system is that non labeled image does not undergo the
process using the CNN model. it only deals with labeled image datasets.

Figure 1: Proposed Architecture

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2907
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue IV Apr 2023- Available at www.ijraset.com

IV. MODULES & THEIR FUNCTIONALITIES

A. Login Module
A user can access an application by providing their username and password or by authenticating with a social media login on the
login screen. To create an Android application, utilize Android Studio and a package. Launch the application, pick an Android device
that is powered on, install it there, and check the outcomes.

B. Registration Module
A registration form is a set of fields that a user fills out and sends to a business or person. You can want someone to complete a
registration form for a variety of reasons. To sign up clients for subscriptions, services, or other programme or plans, businesses
employ registration forms.

C. GUI Module
Instead of text-based UI, typed command labels, or text navigation, a graphical user interface enables users to engage with electronic
devices through graphical icons and auditory indicators like primary notation.

D. Database Module
All user data is gathered using database modules. Users' whole data is stored in the database module when they register for this
application. This data is used to complete the login procedure.

E. Select Image Module

This model is used to choose the gear image that the user will use to determine whether or not the machine's gears are defective. they
accept image as input.

F. CNN process Module

The CNN Algorithm is used in the CNN model. The CNN algorithm takes an image from a chosen image model, processes it, and
outputs the results.

V. CONCLUSION
In this proposed work, gear defect detection system based on machine gear system.It is mainly useful in industrial machine where
machines have more complexity in gear, by using this system we easily find the detected gear and it helps to improve scope ,
efficiency quality and reliability of industrial inspection and it also reduces the bulky maintenance cost in machine industry.

REFERENCES
[1] Gear Assembly Fault Analysis Using Dual-Tree Complex Wavelet Transform Deepika S. Kashawan, K.R.[2016]
[2] A Bevel Gear Quality Inspection System Based on Multi-Camera Vision Technology. Ruiling Liu1, Dexing Zhong2, Hongqiang Lyu3 and Jiuqiang Han4[2012]
[3] Experiment research on the Metal Magnetic Memory in gear micro crack detection. Chenhui Kang1, Lingli Cui2, Jianyu Zhang3, Lixin Gao4, and Yonggang
Xu5[2015]
[4] A Defect Status Detecting Method for External Gear in Railway. Chuangang Wang1, Fuqiang Li2, Yanfeng Li3, Houjin Chen4, Xuyang Cao5[2014]

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2908