Perancangan Sistem Monitoring Kinerja Juru Las pada Aktivitas Pengelasan Pipa Berbasis Wearable Device

Murdwicahyo, Edwin (2023) Perancangan Sistem Monitoring Kinerja Juru Las pada Aktivitas Pengelasan Pipa Berbasis Wearable Device. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 04111940000072-Undergraduate_Thesis.pdf - Accepted Version Restricted to Registered users only until 1 September 2025. Download (14MB) | Request a copy

Abstract

Pengelasan pipa merupakan proses penting dalam industri konstruksi dan manufaktur, dan kualitas pengelasan sangat mempengaruhi keamanan dan keandalan struktur pipa. Oleh karena itu, perancangan sistem monitoring kinerja juru las menjadi krusial dalam memastikan kualitas pengelasan yang baik. Dalam penelitian ini, penulis menggunakan sensor wearable device sebagai alat monitoring untuk merekam data sensorik selama aktivitas pengelasan pipa. Data sensorik yang dihasilkan oleh wearable device, seperti accelerometer, gyroscope, dan magnetometer, digunakan sebagai input untuk sistem pengolahan data. Penulis melakukan pengolahan data menggunakan metode machine learning dan deep learning untuk mengenali aktivitas pengelasan dan mengevaluasi performa juru las. Proses pengolahan data melibatkan langkah-langkah seperti pre-prosessing data, ekstraksi fitur, dan pelatihan model algoritma. Penerapan machine learning dilakukan dengan menggunakan algoritma Support Vector Machine (SVM), sedangkan deep learning menggunakan Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM). Hasil dari kedua metode tersebut dianalisis dan dibandingkan untuk mengevaluasi kinerja sistem. Melalui pelatihan dan pengujian menggunakan SVM dan CNN-LSTM, penulis berhasil mengklasifikasikan aktivitas pengelasan dengan tingkat akurasi yang baik. Hasil analisis menunjukkan bahwa penggunaan deep learning, khususnya dengan menggunakan model CNN-LSTM, menghasilkan akurasi yang lebih baik dalam mengenali dan memprediksi aktivitas pengelasan dibandingkan dengan machine learning. Hasil pelatihan data aktivitas pengelasan melalui SVM dengan beberapa variasi data memiliki nilai akurasi sebesar >98%, sedangkan pelatihan data kinerja juru las melalui CNN-LSTM dengan beberapa kriteria data memiliki nilai akurasi sebesar >99%.
===================================================================================================================================
Pipe welding is a crucial process in the construction and manufacturing industries, and the quality of welding significantly impacts the safety and reliability of the pipe structure. Therefore, designing a performance monitoring system for welders becomes crucial to ensure good welding quality. In this research, the authors utilized wearable devices with sensors as monitoring tools to record sensory data during welding pipe activities. Sensor data generated by wearable devices, such as accelerometers, gyroscopes, and magnetometers, were used as inputs for data processing. The authors employed machine learning and deep learning methods to recognize welding activities and evaluate the welder's performance. The data processing involved steps like data preprocessing, feature extraction, and training algorithm models. Machine learning was applied using the Support Vector Machine (SVM) algorithm, while deep learning used the Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM) model. The results of both methods were analyzed and compared to evaluate the system's performance. Through training and testing with SVM and CNN-LSTM, the authors achieved good accuracy in classifying welding activities. The analysis results indicated that deep learning, especially using the CNN-LSTM model, yielded better accuracy in recognizing and predicting welding activities compared to the machine learning. The training data for welding activities through SVM with various data variations achieved an accuracy rate of >98%, while the training data for the welder's performance using CNN-LSTM with multiple data criteria achieved an accuracy rate of >99%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Pengelasan pipa, Wearable device, Machine learning, Deep learning, Kinerja, Support Vector Machine (SVM); Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM), Pipe welding, Wearable device, Sensor data, Machine learning, Deep learning, Permormance, Support Vector Machine (SVM), Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM).
Subjects:	V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM298.5 Shipbuilding industri. Shipyards
Divisions:	Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User:	Edwin Murdwicahyo
Date Deposited:	18 Aug 2023 03:45
Last Modified:	18 Aug 2023 03:45
URI:	http://repository.its.ac.id/id/eprint/104188

Actions (login required)

View Item