Ta'lim innovatsiyasi va integratsiyasi
47-son_3-to’plam_Iyun -2025
242
ISSN:3030-3621
CREATION OF A SYSTEM FOR PREDICTING FAULTS IN SEWING
EQUIPMENT BASED ON MACHINE LEARNING
Ibroximova Maftunaxon Nozimjon qizi
PhD student, Namangan State Technical University
ibroximovamaftunaxon7@gmail.com
Annotation.
In the process of development of modern industry, ensuring the
uninterrupted operation of production systems, increasing the reliability of equipment
and reducing maintenance costs is becoming increasingly important. Especially in the
light industry, in particular in the garment industry, since each element of the
production line is closely interconnected, the failure of one piece of equipment can stop
the entire production process. Therefore, the enrichment of technical service systems
with innovative approaches is an urgent issue.
Keywords.
Sewing machine, sewing equipment, sewing, industry.
MASHINA O‘RGANISH ASOSIDA TIKUV USKUNALARINING
NOSOZLIKLARINI OLDINDAN BASHORAT QILISH TIZIMINI
YARATISH
Ibroximova Maftunaxon Nozimjon qizi
Tayanch doktorant, Namangan davlat texnika universiteti
ibroximovamaftunaxon7@gmail.com
Annotatsiya.
Zamonaviy sanoatning rivojlanishi jarayonida ishlab chiqarish
tizimlarining uzluksiz ishlashini ta’minlash, uskunalar ishonchliligini oshirish va
xizmat ko‘rsatish xarajatlarini kamaytirish muhim ahamiyat kasb etmoqda. Ayniqsa,
yengil sanoat sohasi, xususan tikuvchilik sanoatida, ishlab chiqarish liniyasining har
bir elementi o‘zaro chambarchas bog‘liq bo‘lgani sababli, bitta uskuna nosozligi butun
ishlab chiqarish jarayonini to‘xtatib qo‘yishi mumkin. Shu bois, texnik xizmat
ko‘rsatish tizimlarining innovatsion yondashuvlar bilan boyitilishi dolzarb masala
hisoblanadi.
Kalit so‘zlar.
Tikuv mashinasi, tikuv uskunalari, tikuvchilik, sanoat.
СОЗДАНИЕ СИСТЕМЫ ПРОГНОЗИРОВАНИЯ НЕИСПРАВНОСТЕЙ
ШВЕЙНОГО ОБОРУДОВАНИЯ НА ОСНОВЕ МАШИННОГО
ОБУЧЕНИЯ
Иброхимова Мафтунахон Нозимжон кизи
Ta'lim innovatsiyasi va integratsiyasi
47-son_3-to’plam_Iyun -2025
243
ISSN:3030-3621
Базовый докторант, Наманганский
государственный технический университет
ibroximovamaftunaxon7@gmail.com
Аннотация.
В процессе развития современной промышленности все
большую значимость приобретает обеспечение бесперебойной работы
производственных систем, повышение надежности оборудования и снижение
затрат на его обслуживание. Особенно в легкой промышленности, в частности в
швейной промышленности, поскольку каждый элемент производственной линии
тесно взаимосвязан, выход из строя одной единицы оборудования может
остановить весь производственный процесс. Поэтому обогащение систем
технического сервиса инновационными подходами является актуальным
вопросом.
Ключевые слова.
Швейная машина, швейное оборудование, шитье,
промышленность.
This study addresses the issue of developing an effective model based on
machine learning algorithms for continuous monitoring of the condition of sewing
equipment, early detection of failures based on collected sensor data, and their
elimination. Traditional maintenance methods (plan-based and reactive approaches)
cannot fully meet modern requirements, as they are performed after a failure occurs or
based on an estimated time frame. Such approaches can be technically unreliable and
also lead to inefficient costs. Therefore, the creation of predictive maintenance systems
using artificial intelligence, especially machine learning (ML) technologies, has
become an important direction.
Although there have been many studies on Predictive Maintenance (PdM)
systems worldwide, there have been relatively few studies on garment industry-specific
failure analysis and prediction using machine learning. These technologies have been
widely implemented in industries such as mechanical engineering, automotive, and
energy. For example, in the intelligent monitoring systems developed by Lee et al.
(2014), the condition of equipment is continuously monitored using sensors and the
probability of failure is determined using artificial neural networks.
In the garment industry, parameters such as vibration, temperature, and rotation
speed collected by sensors can be used to assess equipment condition and predict
problems. While some studies in this field have proposed optimized maintenance
schedules based on machine learning, there is little research on specific equipment
analysis and models suitable for real production. Therefore, this paper aims to analyze
the types of failures specific to specific equipment in the garment industry, collect the
Ta'lim innovatsiyasi va integratsiyasi
47-son_3-to’plam_Iyun -2025
244
ISSN:3030-3621
necessary sensor data, and analyse them based on artificial intelligence to propose an
effective prediction system.
System architecture. The sewing equipment failure prediction system consists of
several key components. Each of them determines the efficiency of the system and is
important for the successful operation of the machine learning model. Data collection
module. Real-time data on the technical condition of sewing equipment is collected
through sensors. The most common malfunctions can be detected by:
• Vibration sensor – to detect uneven movement and imbalance of the internal
mechanisms of the equipment.
• Temperature sensor – to monitor the overheating of the engine or other moving
parts.
• Rotational speed sensor (RPM) – to detect changes in the operation of the motors.
• Sound sensor (acoustic monitoring) – to detect signs of malfunction by the
appearance of strange sounds during operation.
The data collected from the sensors is transmitted to a central database via
microcontrollers (e.g. Raspberry Pi or Arduino). MQTT or Wi-Fi technologies can be
used for data transfer. Data pre-processing. The raw data obtained is processed through
the following steps: Cleaning: Identify and correct incorrect or missing values;
Normalization: Bring values to unity to eliminate differences between sensors;
Windowing: Divide data into small intervals for time-based analysis; Feature
Extraction: Extract statistical indicators (mean, variance, vibration level).
Model training and testing. Machine learning algorithms are used to predict
sewing equipment failures. The following algorithms were compared in the study.
Below you will find the selected algorithms:
Random Forest – classification based on a set of unrelated trees.
Support Vector Machine (SVM) – for classification problems with complex
boundaries.
Gradient Boosting (XGBoost) – for cases where high accuracy is required.
LSTM (Long Short-Term Memory) – for predictions based on time series (if
real-time analysis is required).
Training and testing process: Dataset creation: A data set is formed, collected by
sensors and with predefined fault states (labels); Train/Test split: Usually 70% is
allocated for training, 30% for testing; Model training: Algorithms learn, that is, they
learn to predict whether the device is healthy or faulty based on various features;
Accuracy indicators: confusion matrix, precision, Recall, F1-score and ROC-AUC
curve (if necessary).
Initial results showed that the Random Forest algorithm worked with high
accuracy. It was found that vibration and temperature parameters play a key role in the
model. The LSTM algorithm was useful in detecting changes in the real-time flow, but
Ta'lim innovatsiyasi va integratsiyasi
47-son_3-to’plam_Iyun -2025
245
ISSN:3030-3621
required more computational resources. The prediction system developed in this study
was tested on the basis of experimental equipment close to real industrial conditions.
Approximately 10,000 data samples of equipment operation in different conditions
were collected based on 3 weeks of monitoring data collected using sensors. The
following machine learning algorithms were compared with each other and the
following performance indicators were recorded:
Algorithm
Accuracy
F1-score ROC-AUC
Random Forest
94.7%
0.93
0.96
SVM
89.3%
0.87
0.91
XGBoost
95.1%
0.94
0.97
LSTM (Recurrent NN) 92.5%
0.91
0.95
The XGBoost algorithm also demonstrated the highest accuracy and stability.
However, in cases where real-time performance was required, the LSTM algorithm
showed advantages, as it allowed for deeper analysis of time-varying patterns.
Feature Importance. Using the Random Forest and XGBoost algorithms,
important features were identified, and the following were identified as the most
important input parameters of the model:
• Average value and variance of vibration variation
• Motor temperature variation
• Rotational speed fluctuations
• Changes in the sound spectrum during operation
These features made it possible to determine with high reliability whether the
equipment is healthy or faulty. This study investigated the possibilities of early
detection of sewing equipment failures using machine-learning technologies.
Experiments showed that: Based on real-time data collected by sensors, failures can be
reliably predicted; Boost and Random Forest algorithms work with high accuracy and
are suitable for industrial conditions; such systems can optimize the maintenance
schedule, increase equipment reliability, and prevent production interruptions.
Integrate the system with a wider sensor base (e.g. pressure, current sensors).
Expand the database and create a deep model based on Deep Learning. Launch a pilot
project at real industrial enterprises and assess economic efficiency. Develop a mobile
application or web platform for providing services based on artificial intelligence.
REFERENCES
1.
Liu, X., et al. (2017). "Predictive Maintenance of Industrial Equipment Using
Machine Learning: A Case Study of Manufacturing."
Journal of Manufacturing
Science and Engineering
.
2.
Jia, F., et al. (2019). "Machine learning for predictive maintenance: A systematic
review and perspective."
Computers & Industrial Engineering
.
Ta'lim innovatsiyasi va integratsiyasi
47-son_3-to’plam_Iyun -2025
246
ISSN:3030-3621
3.
Vaidyanathan, V., et al. (2021). "Anomaly detection in industrial systems for
predictive maintenance using machine learning techniques."
Proceedings of the
IEEE International Conference on Industrial Technology
.
4.
Zhao, X., et al. (2020). "Deep Learning-Based Prognostics for Predictive
Maintenance in Manufacturing."
IEEE Transactions on Industrial Informatics
.
5.
Iyer, S., & Tan, C. Y. (2020). "Data-Driven Predictive Maintenance: A Review and
Case Study."
International Journal of Advanced Manufacturing Technology
.
6.
Zheng, J., et al. (2022). "Application of Machine Learning for Predictive
Maintenance in Textile Industry."
Procedia CIRP
.
7.
Salah, A., & Karray, F. (2019). "Predictive maintenance of industrial machines
using machine learning algorithms."
Journal of Industrial Information Integration
.
