Analisis Perbandingan Prediksi Kesehatan Diesel Engine Menggunakan Support Vector Machine(SVM) Dan Back Propagation Neural Network (BPNN)

Nurdin, Fadli (2025) Analisis Perbandingan Prediksi Kesehatan Diesel Engine Menggunakan Support Vector Machine(SVM) Dan Back Propagation Neural Network (BPNN). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6007221027-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Condition Based Predictive Maintenance (CBM) merupakan pendekatan penting untuk meningkatkan keandalan mesin diesel. Penelitian ini membandingkan Support Vector Machine (SVM) yang dimodifikasi dengan Back Propagation Neural Network (BPNN) dalam memprediksi kesehatan mesin diesel berdasarkan parameter operasional antara lain RPM mesin, tekanan oli pelumas, tekanan bahan bakar, tekanan pendingin, suhu oli mesin dan suhu pendinginan. Analisis korelasi Spearman menunjukkan bahwa tiga parameter operasional yaitu RPM mesin (r = 0.2543), tekanan oli pelumas (r = 0.1604) dan suhu pendingin (r = 0.1764) memiliki pengaruh terbesar terhadap kesehatan mesin. Uji ANOVA Deviance dengan model regresi logistik digunakan karena menyesuaikan jenis data kondisi mesin berupa data biner 1 (sehat) dan 0 (tidak sehat) mengonfirmasi bahwa RPM mesin (p-value = 0.002) dan suhu pendinginan (p-value = 0.01) adalah faktor paling signifikan dalam prediksi kondisi mesin (p-value < 0.05). ANOVA Deviance yang digunakan valid berdasarkan uji Goodness-of-Fit (p-value > 0.05). Dalam pemodelan prediksi, penelitian sebelumnya menggunakan SVM dengan Linear dan Polynomial Kernel, namun hasilnya kurang optimal karena tingkat akurasi yang masih rendah dan kesalahan klasifikasi yang tinggi, terutama dalam mendeteksi kondisi mesin tidak sehat. Untuk mengatasi keterbatasan ini, penelitian ini mengembangkan SVM dengan kernel Sigmoid dan RBF serta menggunakan model lai yaitu BPNN. SVM dengan kernel RBF mencapai akurasi 85.46% namun yang tertinggi didapatkan pada BPNN dengan konfigurasi [2 2 1] menggunakan fungsi aktivasi satlins mencapai akurasi 97.16%, menunjukkan keunggulan dalam menangkap pola kesehatan mesin. Hasil penelitian ini menegaskan bahwa BPNN lebih akurat dibandingkan metode sebelumnya. Fokus pemantauan CBM pada tekanan pendingin dan tekanan oli pelumas dapat meningkatkan efisiensi pemeliharaan serta mengurangi risiko kegagalan mesin diesel.
===================================================================================================================================
Condition-Based Predictive Maintenance (CBM) is a crucial approach to enhancing the reliability of diesel engines. This study compares a modified Support Vector Machine (SVM) with a Back Propagation Neural Network (BPNN) in predicting diesel engine health based on operational parameters, including engine RPM, lubricating oil pressure, fuel pressure, coolant pressure, oil temperature, and coolant temperature. Spearman correlation analysis revealed that three operational parameters—engine RPM (r = 0.2543), lubricating oil pressure (r = 0.1604), and coolant temperature (r = 0.1764)—had the strongest influence on engine health. ANOVA Deviance testing using logistic regression was employed to accommodate the binary nature of engine condition data (1 = healthy, 0 = unhealthy). The results confirmed that engine RPM (p-value = 0.002) and coolant temperature (p-value = 0.01) were the most significant factors in predicting engine condition (p-value < 0.05). The validity of the ANOVA Deviance analysis was supported by the Goodness-of-Fit test (p-value > 0.05). While previous studies employed SVM with linear and polynomial kernels, their results were suboptimal due to lower accuracy and high misclassification rates, particularly in identifying unhealthy engine conditions. To overcome these limitations, this study developed SVM models with Sigmoid and RBF kernels and introduced BPNN as an alternative approach. The SVM with RBF kernel achieved an accuracy of 85.46%, while the highest performance was observed in the BPNN with a [2 2 1] configuration using the satlins activation function, reaching 97.16% accuracy, demonstrating superior capability in capturing engine health patterns. These findings affirm that BPNN outperforms previous methods in accuracy. Emphasizing CBM monitoring on coolant temperature and lubricating oil pressure can enhance maintenance efficiency and reduce the risk of diesel engine failure.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Support Vector Machine (SVM), Back Propagation Neural Network (BPNN), Sigmoid Kernel, K-Means Clustering, Korelasi Spearman, ANOVA Deviance, Prediksi Kesehatan Mesin, CBM.; Support Vector Machine (SVM), Back Propagation Neural Network (BPNN), Sigmoid Kernel, K-Means Clustering, Spearman Correlation, ANOVA Deviance, Engine Health Prediction, CBM.
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ174 Maintenance and repair of machinery
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Mechanical Engineering > 21101-(S2) Master Thesis
Depositing User:	Fadli Nurdin
Date Deposited:	06 Aug 2025 01:10
Last Modified:	06 Aug 2025 01:10
URI:	http://repository.its.ac.id/id/eprint/127625

Actions (login required)

View Item