Perancangan Sistem Deteksi Keausan Pahat Pada Mesin Milling Menggunakan "Cubic SVM"

Hikmah, Zahra Putri Nurul (2023) Perancangan Sistem Deteksi Keausan Pahat Pada Mesin Milling Menggunakan "Cubic SVM". Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009211147-Undergraduate_Thesis.pdf Restricted to Repository staff only until 3 October 2025. Download (7MB) | Request a copy

Abstract

Kerusakan suatu mesin dapat terjadi begitu saja tanpa timbulnya suatu indikasi tertentu. Sehingga dibutuhkan suatu sistem monitoring yang dapat digunakan untuk mendeteksi suatu anomali dalam proses permesinan. Deteksi anomali merupakan suatu proses untuk mengidentifikasikan suatu error atau peristiwa yang tidak terduga. Penelitian ini bertujuan untuk mengetahui sinyal vibrasi atau sinyal arus, sinyal yang paling tepat digunakan untuk mendeteksi keausan pahat pada mesin milling 3 axis Matsura 510 menggunakan model machine learning cubic SVM dan juga untuk mengetahui pengaruh penggunaan dekomposisi sinyal EMD terhadap hasil klasifikasi data yang dihasilkan. Penelitian dimulai dari pengambilan data untuk mendapatkan sinyal arus dan sinyal vibrasi kemudian masing-masing sinyal di-prepocessing. Lalu setelah itu sinyal diubah menjadi numerical features melalui proses features extraction. Numerical features yang sudah didapatkan kemudian diseleksi menggunakan ANOVA supaya menghasilkan hasil performance metrics yang baik. Sebelum features diklasifikasikan dengan model machine learning, maka diperlukan validasi model untuk menghindari overfitting dan underfitting. Setelah itu features dapat diklasifikasikan dengan menggunakan dua proses yaitu proses training dan proses testing sehingga didapatkan prediksi mengenai kondisi mesin. Berdasarkan penelitian yang telah dilakukan dapat disimpulkan bahwa ternyata sinyal vibrasi dalam domain frekuensi yang diolah tanpa menggunakan EMD dapat digunakan untuk mendeteksi keasuan pahat pada mesin milling 3 axis Matsura 510. Hal tersebut dapat dibuktikan dengan nilai precision dan recall yang dihasilkan dari proses training dan testing. Pada proses training didapatkan nilai precision sebesar 94,7% dan recall sebesar 84,7%. Kemudian pada proses testing didapatkan nilai precision sebesar 90,1% dan recall sebesar 90,1%.
==============================================================================================================================
Damage to a machine can occur without any indication. Therefore require monitoring system can be used to detect an anomaly in the machining process. Anomaly detection is a process to identify an error or unexpected event. This research aims to find out between vibration signals and current signals, which signals are the best to detect tool wear on Matsura 510 3-axis milling machine using a cubic SVM machine learning algorithm. In addition, this research also aims to determine the effect of using EMD signal decomposition on the resulting data classification results. The research begins with data collection to obtain current signals and vibration signals, then each signal is preprocessed. Then after that, the signal is converted into numerical features through the features extraction process. Numerical features that have been obtained are then selected using ANOVA to produce good performance metrics. Before features are classified with a machine learning model, it is necessary to validate the model to avoid overfitting and underfitting. After that, the features can be classified using two processes, the training process and the testing process so that predictions about the condition of the machine can be obtained. Based on the research that has been done, it would be concluded that it turns out that the vibration signal processed in frequency domain without using EMD can be used to detect the acidity of the tool on the Matsura 510 3-axis milling machine. That could be proved by the precision and recall values produced in training and testing data. In the training process, the precision value having a percentage of 94,7% and 84,7%. Then in the testing process obtained precision value of 90,1% and recall of 90,1%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Cubic SVM, Deteksi anomali, Anomaly detection, EMD, Machine learning.
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Zahra Putri Nurul Hikmah
Date Deposited:	31 Jul 2023 02:18
Last Modified:	31 Jul 2023 02:18
URI:	http://repository.its.ac.id/id/eprint/100911

Actions (login required)

View Item