LEE- Deniz Ulaştırma Mühendisliği-Doktora

Bu koleksiyon için kalıcı URI

http://hdl.handle.net/11527/19252

Gözat

Konu "bilgisayar destekli bakım" ile LEE- Deniz Ulaştırma Mühendisliği-Doktora'a göz atma
  • Öge
    A prescriptive analytics approach towards critical ship machinery operations
    (Graduate School, 2024-07-09) Yiğin, Barış ; Çelik, Metin ; 512182001 ; Maritime Transportation Engineering
    Shipping handles more than 70% of global trade, is a pillar of the supply chain. To ensure safe, reliable, and environmentally responsible operatins, shipowners and operators must maintain their vessels' operational status at all times. Maintenance standards are essential for keeping both main and auxiliary machinery in optimal condition, thereby ensuring reliable and safe operations. These programs aim to maintain high performance with minimal impact on service, recognizing that the cost savings from effective maintenance program can prevent drawbacks due to machinery faults. The main objective of machine maintenance is to maximize availability by extending the service life of ship machineries and eliminating potential failures by early detection. This involves finding the finding the optimum maintenance strategy, as even minor failure can cause irreversible damage to the entire system if not promptly addressed. Given the complexity and interdependencies within marine systems, a proactive maintenance approach is crucial. Due to scarcity of labeled data and anomalous data, the research question of anomaly detection always attracted interest from academia and industry. Implementing anomaly detection technologies is a challenging task in marine systems due to their complexities and external factors. To address these challenges, this study proposes a prescriptive analytics framework that combines predictive analytics and decision support systems. This framework leverages data collected from various sensors installed on ship machinery to monitor performance and detect anomalies. One of the key innovations of this research it employes data augmentation techniques to generate realistic synthetic failure data, further enhancing the robustness of predictive models. The implementation of this prescriptive maintenance framework involves several steps. First, a comprehensive Failure Mode and Effect Analysis (FMEA) is conducted to identify potential failure modes, cause of failures and effects of failures. This analysis helps prioritization of the maintenance activities based on the criticality of different failure modes. Next, with the use of data augmentation technique called Generative Adversarial Network, synthetic data generation carried out to create faulty data information. This faulty data generation step enhance the training pool before the next step of anomaly detection process. In order to perform anomaly detection, six different classifiers namely, logistic regression, decision trees, random forest, K nearest neighbor, AdaBoost and XGBoost algorithms trained and validated using historical data and the generated synthetic data. Data set used in this study includes real time data collected from field on a diesel generator installed on a 310,000 DWT oil tanker. The field data collection took place over a six month period and it includes 33 features and 259,200 row data. Findings from the study yield promising results achieving 83.13% accuracy with use XGBoost algorithm and other ranging between 67% to 81%. Finally, a decision support system is integrated to provide actionable recommendations to ship operators, optimizing maintenance schedules and resource allocation. The results of the field study conducted as part of this research demonstrate the effectiveness of the proposed framework. Ships equipped with the prescriptive maintenance system has a significant potential for reduction in unexpected machinery failures, maintenance cost and autonomy of decision in case of anomalies. The system also offers improvement of overall operational efficiency and reliability of the ships. In conclusion, prescriptive maintenance the pinnacle of modern maintenance strategies, offering returns in terms of equipment reliability, safety and operational efficiency. Although installation of data acquisition systems may require initial investment, the benefits include reduced operational distruption and optimized maintenance budgets, making it a valuable approach for the maritime industry.