Abstract: Sensor fault detection under marine engine centered localized models of an engine propeller combinator diagram is presented in this study. The proposed approach consists of two detection levels to identify of sensor fault situations in an onboard data acquisition system of a vessel. Each parameter in ship performance and navigation data can have a realistic data range (i.e. a threshold relates to the variance), where the parameter can vary. If the sensor reads a value beyond this parameter range, then that data point is categorized as a sensor fault situation by the first fault detection level. However, some sensor faults are located within this data range and that cannot identify by this detection level. Such complex sensor fault situations are detected by the second fault detection level by considering the proposed localized models. These localized models are derived with respect to the operating regions of an engine-propeller combinator diagram, where the respective data points are clustered by Gaussian mixture models with an expectation maximization algorithm. Each data cluster is examined through principal component analysis and projected into the bottom principal component to identify such complex sensor fault situations. A data set of ship performance and navigation information of a selected vessel is used through these sensor fault detection levels and the successful results on identifying such sensor fault situations are also presented in this study.
[1]
Lokukaluge P. Perera,et al.
Identification of Optimal Trim Configurations to improve Energy Efficiency in Ships
,
2015
.
[2]
Linear Autoencoder Networks for Structured Data
,
2013
.
[3]
Ulrik Dam Nielsen,et al.
FAULT DETECTION FOR SHIPBOARD MONITORING AND DECISION SUPPORT SYSTEMS
,
2009
.
[4]
Ulrik Dam Nielsen,et al.
Fault Detection for Shipboard Monitoring Volterra Kernel and Hammerstein Model Approaches
,
2009
.
[5]
Lokukaluge P. Perera,et al.
Ship speed power performance under relative wind profiles
,
2016
.
[6]
Lokukaluge P. Perera,et al.
Data Compression of Ship Performance and Navigation Information Under Deep Learning
,
2016
.
[7]
Lokukaluge P. Perera,et al.
Data Analytics for Capturing Marine Engine Operating Regions for Ship Performance Monitoring
,
2016
.
[8]
Lokukaluge P. Perera,et al.
Evaluations on Ship Performance Under Varying Operational Conditions
,
2015
.
[9]
Lokukaluge P. Perera,et al.
Machine intelligence for energy efficient ships: A big data solution
,
2016
.