Pengembangan Model Hybrid Predictive Method Untuk Diagnostic Assessment Pada Mesin Kapal Berdasarkan Data Condition Monitoring

Siswantoro, Nurhadi (2020) Pengembangan Model Hybrid Predictive Method Untuk Diagnostic Assessment Pada Mesin Kapal Berdasarkan Data Condition Monitoring. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
04111760010006-Disertation.pdf - Accepted Version
Restricted to Repository staff only

Download (8MB) | Request a copy

Abstract

Mesin diesel adalah jenis mesin pembakaran dalam (internal) di mana bahan bakar dinyalakan oleh suhu tinggi gas yang dikompresi. Untuk motor penggerak di kapal, banyak digunakan mesin diesel dengan kategori marine diesel. Untuk memenuhi kerja mesin diesel di kapal terdapat beberapa sistem yang harus diperhatikan, seperti: sistem bahan bakar (fuel oil system), sistem pelumas (lubricating oil system), sistem pendingin (cooling system), dan starting air system. Kerusakan-kerusakan yang terjadi pada mesin kapal akan berakibat fatal apabila tidak dilakukan kegiatan monitoring dan perawatan yang baik pada saat kapal dalam kondisi operasi. Selain itu, pola kegagalan pada peralatan generasi keempat secara statistik 80% berhubungan dengan random failure, maka kegiatan predictive maintenance menjadi sangat penting. Bilamana predictive maintenance diterapkan pada semua peralatan di keseluruhan sistem tersebut, tentu akan sangat tidak ekonomis. Sehingga perlu ada sebuah metode ilmiah untuk melakukan penentuan prioritas perawatan yang menentukan mana saja equipment diperlukan kegiatan predictive maintenance. Strategi perawatan yang mengarah ke predictive maintenance juga telah dimulai penerapannya oleh beberapa badan klasifikasi (classification society), seperti Lloyd Register, Nippon Kaiji Kyokai, DNV-GL, dan American Bureau of Shipping. Konsep perawatan dengan menerapkan criticality assessment dan condition monitoring telah banyak dilakukan untuk menentukan level risiko dan tindakan perawatan. Namun konsep criticality assessment dan condition monitoring yang dilakukan belum satu kesatuan yang terintegrasi, yang mana berhenti pada satu assessment berupa criticality saja atau condition monitoring saja. Urgensi tersebut memerlukan solusi berupa sebuah metode yang terintegrasi. Hybrid Predictive Method (HPM) mengakomodasi penentuan prioritas perawatan dengan pendekatan criticality assessment guna menentukan prioritas predictive assessment. Konsep Hybrid Predictive Method (HPM) yang diusulkan merupakan konsep mengombinasikan reliability tools dan predictive/forecasting tools. Reliability tools untuk criticality assessment adalah dengan Failure Mode and Effect Criticality Analysis (FMECA) berbasis pendekatan fuzzy logic. Pendekatan bottom-up approach FMECA dimaksudkan untuk menggali failure modes yang memberikan potensi kegagalan pada main engine system. Failure modes ini kemudian dilakukan analisis untuk penentuan level kekritisan equipment. Teori fuzzy logic yang ditambahkan pada FMECA menampung ketidakpastian akibat samarnya informasi yang dimiliki maupun unsur preferensi yang subjektif yang digunakan dalam penilaian terhadap mode kegagalan yang terjadi. Proses predictive asessment menggunakan pendekatan Multilayer Perceptron (MLP) dengan metode Artificial Neural Network (ANN). ANN memiliki kelebihan untuk self-learning, adaptivity, fault tolerance, nonlinearity, and advancement in input to an output mapping. Algoritma Artificial Neural Network berbasis Multilayer perceptron (MLP) didasarkan pada hubungan jika output memberikan hasil yang salah, maka penimbang (bobot) dikoreksi supaya galatnya dapat diperkecil dan respon jaringan selanjutnya diharapkan akan lebih mendekati target. Multilayer perceptron juga berkemampuan untuk memperbaiki penimbang pada lapisan tersembunyi (hidden layer). Hasil pemodelan Hybrid Predictive Method yang pertama adalah berupa luaran hasil analisis FMECA fuzzy. Hasil penelitian menunjukkan terdapat 78 equipment yang terbagi dalam main engine, fuel oil system, lubricating oil system, cooling system dan starting air system dengan total failure mode sejumlah 339. Dari keseluruhan failure mode, apabila dilihat sebaran level kekritisan memiliki level low risk (23%), medium risk (67%), dan high risk (10%). Level high risk didapatkan pada failure modes yang memiliki nilai fuzzy risk priority number (FRPN) 60-100. Sedangkan pada level medium risk nilai FRPN adalah berkisar antara 40-60, dan level low risk memiliki nilai FRPN 0-40. Hasil penilaian FMECA fuzzy menjadi dasar untuk menentukan prioritas perawatan berupa condition monitoring dan predictive assessment pada sebuah peralatan yang mengalami kegagalan pada kondisi high risk. Condition monitoring dan prediksi yang telah dilakukan berpengaruh terhadap diagnostic assessment pada main engine. Hasil diagnosis assessment saat ini, menunjukkn kondisi main engine dalam keadaan masih normal. Namun trending prediksi temperatur gas buang menunjukkan kenaikan, combustion dan compression pressure yang menurun perlu dipersiapkan untuk penentuan jadwal pemeriksaan/ survey. Saat ini regulasi yang berlaku di industri pelayaran Indonesia adalah bersifat time-based maintenance, yang mana terbagi dalam annual survey, intermediate survey, dan special survey. Hybrid Predictive Method (HPM) memiliki kelebihan untuk membuka peluang dalam penerapan perawatan berbasis risiko dan prediksi. Hybrid Predictive Method dapat mengelompokkan failure modes yang high risk dan memprediksi waktu parameter menyentuh batas technical operation yang diijinkan. Batas technical operation ini menjadi penentuan waktu penjadwalan tindakan perawatan/ overhaul. Dalam penelitian ini telah divalidasi predictive assessment dengan Artificial Neural Network berbasis Multilayer Perceptron (MLP) memiliki error kurang dari 5%. Dengan demikian hasilnya akurat untuk memprediksi periode selanjutnya. ======================================================================================================================== A diesel engine is a type of internal combustion engine. Marine diesel engine is widely used as prime movers on ships. In order to ensure the diesel engine works properly, one of the considerations is the condition of the equipment on main engine system. The equipment should have high reliability, availability, and good performance conditions. Main engine system consist of fuel oil system, lubricating oil system, cooling system and starting air system. Failures that occur in the ship engine will be fatal if monitoring activities are not carried out and proper maintenance when the ship is in operational condition. In addition, the failure pattern of the fourth generation equipment is statistically 80% related to random failure, so predictive maintenance is very important. If predictive maintenance is applied to all equipment in the entire system, it will certainly be very uneconomical. It needs a scientific method to determine the priority of maintenance on the equipment that required for predictive maintenance activities. Several classification societies have also begun implementing a maintenance strategy that leads to predictive maintenance, such as the Lloyd Register, Nippon Kaiji Kyokai, DNV-GL, and the American Bureau of Shipping. The concept of maintenance by applying criticality assessment and condition monitoring has been carried out to determine the level of risk and maintenace action. However, the concept of criticality assessment and condition monitoring carried out is unintegrated, which in the form of criticality or condition monitoring only. This urgency requires a solution in the form of an integrated method. Hybrid Predictive Method (HPM) accommodates prioritization of maintenance using a criticality assessment approach to determine the priority of predictive assessment. The proposed Hybrid Predictive Method (HPM) concept combines reliability tool and predictive/ forecasting tool. Reliability tool for criticality assessment is the Failure Mode and Effect Criticality Analysis (FMECA) based on the fuzzy logic approach. FMECA's bottom-up approach is intended to explore failure modes that provide potential failure in the main engine system. These failure modes are then analyzed to determine the criticality level of the equipment. The fuzzy logic theory added to FMECA accommodates uncertainty due to obscure information as well as subjective preference elements that are used in the assessment of failure modes. The predictive assessment process uses the Multilayer Perceptron (MLP) approach using the Artificial Neural Network (ANN) method. ANN has advantages for self-learning, adaptivity, fault tolerance, nonlinearity, and advancement in input to an output mapping. Multilayer perceptron-based Artificial Neural Network (MLP) algorithm is based on the relationship if the output gives the false result, then the weight is corrected so that the error can be minimized and the network response is then expected to be closer to the target. Multilayer perceptron also has the ability to fix the weighing on the hidden layer. The first result of the Hybrid Predictive Method modeling is the outputs of the fuzzy FMECA analysis. The results showed that there were 78 equipment divided into main engine, fuel oil system, lubricating oil system, cooling system and starting air system with total failure modes of 339. From the overall failure modes, the critical level distributions have low risk level (23%), medium risk (67%), and high risk (10%). The high risk level is obtained in failure modes which have a fuzzy risk priority number (FRPN) score of 60-100. Whereas at the medium risk level, the FRPN score of 40-60, and the low risk level has a FRPN score of 0-40. The results of the fuzzy FMECA assessment become the basis to determine the maintenance priority in the form of condition monitoring and predictive assessment of an equipment that has high risk level. Condition monitoring and prediction affect the diagnostic assessment on the main engine. The results of the current diagnostic assessment indicate the condition of the main engine is still normal. However, the trending of exhaust gas temperature prediction shows an increase, combustion and compression pressure which shows a decrease need to be prepared for determining the inspection/survey schedule. The regulations in the Indonesian shipping industry are time-based maintenance, which is divided into annual, intermediate and special surveys. Therefore, Hybrid Predictive Method (HPM) has the advantage to cultivate opportunities of predictive and risk-based maintenance application. Hybrid Predictive Method can classify high risk failure modes and predict the time when the parameters reach the allowable technical operation limit. This technical operation limit determines the time for scheduling maintenance/overhaul actions. In this research, predictive assessment using an Artificial Neural Network based on Multilayer Perceptron (MLP) has been validated with an error of less than 5%. Thus the results are accurate to predict the next period.

Item Type: Thesis (Doctoral)
Uncontrolled Keywords: Diagnostic Assessment, FMECA, Fuzzy, Hybrid Predictive Method, Diagnostic Assessment, FMECA, Fuzzy, Hybrid Predictive Method.
Subjects: V Naval Science > VC Naval Maintenance
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) > Ocean Engineering > 38001-(S3) PhD Thesis
Depositing User: Nurhadi Siswantoro
Date Deposited: 24 Aug 2020 04:10
Last Modified: 24 Aug 2020 04:10
URI: https://repository.its.ac.id/id/eprint/80292

Actions (login required)

View Item View Item