Diesel Engine Predictive Performance Using Artificial Neural Network

Rasheesaa, Nadira Afra (2023) Diesel Engine Predictive Performance Using Artificial Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 04211941000011-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (8MB) | Request a copy

Abstract

The combination of available digitalization technologies and the need for monitoring to maximize the performance of crucial equipment on ships without compromising safety has led to numerous studies exploring the utilization of artificial intelligence. In this context, the implementation of machine learning, a form of artificial intelligence, will be applied to two defined objects: ship main engine and diesel engine. The machine learning model for forecasting developed using the Artificial Neural Network (ANN) algorithm will be tested and evaluated in terms of its performance in predicting the exhaust gas temperature for both cases. In this study, the formation of the ANN model will employ the Neural Network Time Series technique, specifically the Nonlinear Autoregressive with Exogenous Inputs (NARX) network type. The expected outcome of this study, when assessing the performance of the ANN model, is to achieve results as close as possible to the actual data. The evaluation of the ANN model will consider the calculation of Mean Squared Error (MSE) and R-squared value, representing the coefficient of correlation between target and actual values. The study generates models tested on each of the 6 cylinders, with MSE ranging from 12.09 to 22.9 and R-squared ranging from 0.66 to 0.78. Meanwhile, in the second case, the study yields an MSE value of 0.058 and an R-squared value of 0.96. Regarding the performance of the ship main engine, the prediction results show fluctuations, with only cylinder 2 slightly exceeding the maximum temperature limit. In the performance analysis, the Artificial Neural Network (ANN) demonstrates patterns and values that closely resemble the actual data. In the historical data, it is observed that cylinder 2 has exceeded the maximum limit temperature on several occasions, the predictions generated by the ANN model indicate the potential for cylinder 2 to experience the same issue again. The predictions for the other 5 cylinders do not show any potential for overheating, as their temperatures have not exceeded the maximum limit since the beginning, given the same engine load.
=================================================================================================================================
Kombinasi teknologi digitalisasi yang tersedia dan kebutuhan pemantauan untuk memaksimalkan kinerja peralatan penting di kapal tanpa mengorbankan keselamatan telah menghasilkan banyak penelitian yang mengeksplorasi pemanfaatan kecerdasan buatan. Dalam konteks ini, implementasi machine learning, salah satu bentuk kecerdasan buatan, akan diterapkan pada dua objek yang ditentukan: mesin utama kapal dan mesin diesel. Model machine learning untuk peramalan yang dikembangkan menggunakan algoritma Artificial Neural Network (ANN) akan diuji dan dievaluasi kinerjanya dalam memprediksi suhu gas buang untuk kedua kasus tersebut. Pada penelitian ini, pembentukan model ANN akan menggunakan teknik Neural Network Time Series, khususnya tipe jaringan Nonlinear Autoregressive with Exogenous Inputs (NARX). Hasil yang diharapkan dari penelitian ini, ketika menilai kinerja model ANN, adalah untuk mencapai hasil yang sedekat mungkin dengan data sebenarnya. Evaluasi model ANN akan mempertimbangkan perhitungan Mean Squared Error (MSE) dan nilai R-squared yang merupakan koefisien korelasi antara nilai target dan nilai aktual. Studi menghasilkan model yang diuji pada masing-masing 6 silinder, dengan MSE berkisar antara 12,09 hingga 22,9 dan R-kuadrat berkisar antara 0,66 hingga 0,78. Sedangkan pada kasus kedua, penelitian menghasilkan nilai MSE sebesar 0,058 dan nilai R-squared sebesar 0,96. Berkaitan dengan performa mesin utama kapal, hasil prediksi menunjukkan fluktuasi, dengan hanya silinder 2 yang sedikit melebihi batas suhu maksimum. Dalam analisis performa, ANN menunjukkan pola dan nilai yang mirip dengan data aktual. Pada data historis, teramati bahwa silinder 2 telah melebihi batas suhu maksimum dalam beberapa kesempatan, dan prediksi yang dihasilkan oleh model ANN menunjukkan potensi bagi silinder 2 mengalami masalah serupa kembali. Prediksi untuk 5 silinder lainnya tidak menunjukkan potensi untuk overheating, karena suhu mereka tidak pernah melebihi batas maksimum sejak awal, meskipun dengan engine load yang sama.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Machine learning, NARX, Forecasting, Diesel Engine
Subjects:	V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM731 Marine Engines
Divisions:	Faculty of Marine Technology (MARTECH) > Marine Engineering > 36202-(S1) Undergraduate Thesis
Depositing User:	Nadira Afra Rasheesa
Date Deposited:	13 Oct 2023 01:51
Last Modified:	13 Oct 2023 01:51
URI:	http://repository.its.ac.id/id/eprint/103570

Actions (login required)

View Item