Improving efficiency of diesel engine fault detection based on multi-source data
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
As the core component of a ship’s engine, the timely and effective maintenance of the cylinder is crucial for the vessel’s stable navigation. Data-driven fault detection methods, which rely solely on historical data from the ship’s systems, are widely used for fault detection. These methods focus on modeling and prediction based exclusively on target data. While previous studies have often employed more complex algorithms and composite models to improve prediction performance, this study introduces a novel approach by integrating a variety of engine room data, breaking through the limitations of traditional methods that rely solely on target data. By fusing the target cylinder data with additional engine room data, a multi-source data collaborative modeling framework was constructed. Building on this, we designed a hybrid model combining deep learning with an Exponentially Weighted Moving Average (EWMA) for real-time monitoring and fault warning of the diesel engine cylinder’s operating status. The experiment compared three time series deep learning algorithms, and the results demonstrated that the collaborative modeling of multi-source data significantly improved prediction accuracy and effectively reduced the average run length (ARL). This innovative method improves the efficiency of fault detection.
Abstract
As the core component of a ship’s engine, the timely and effective maintenance of the cylinder is crucial for the vessel’s stable navigation. Data-driven fault detection methods, which rely solely on historical data from the ship’s systems, are widely used for fault detection. These methods focus on modeling and prediction based exclusively on target data. While previous studies have often employed more complex algorithms and composite models to improve prediction performance, this study introduces a novel approach by integrating a variety of engine room data, breaking through the limitations of traditional methods that rely solely on target data. By fusing the target cylinder data with additional engine room data, a multi-source data collaborative modeling framework was constructed. Building on this, we designed a hybrid model combining deep learning with an Exponentially Weighted Moving Average (EWMA) for real-time monitoring and fault warning of the diesel engine cylinder’s operating status. The experiment compared three time series deep learning algorithms, and the results demonstrated that the collaborative modeling of multi-source data significantly improved prediction accuracy and effectively reduced the average run length (ARL). This innovative method improves the efficiency of fault detection.