Onboard Decision Support Systems

By Zoran Lajic, PhD student at the Technical University of Denmark, Department for Mechanical Engineering.

This project has shown that by using a new approach to increase the reliability of onboard monitoring systems it is possible to get more accurate and dependable data and at the same time reduce the hardware requirements. In the project, the focus has been on wave estimation algorithms used in onboard decision support systems. In this project, the focus has been on the SeaSense system, which identifies critical forthcoming events and gives advice regarding speed and course.

In general, decision support systems on board ships are dependent on different kinds of sensor readings regarding ship motions, waves etc according to their purpose. A potential weakness for a decision support system is the quality of data provided through these readings, especially if the sensor readings are insufficient or absent. Obviously, this weakness decreases the operational value of a decision support system.

In SeaSense, the data is based on the combined use of a mathematical model and measurements from a set of sensors, which monitors relative wave motions,accelerations, green water on deck and structural loading of the hull girder. Should the relative wave motion sensors fail to provide a reliable signal the wave estimator will be equally unreliable.

The basic idea of the project has therefore been to increase the fault tolerance and improve the estimation and prioritising of readings from the different sensors by improving on the underlying mathematical model.

In order to create a fault detection and isolation system that discards faulty sensor readings, the fault diagnosis system was divided into three steps: Fault detection, Fault isolation and Fault identification and estimation.

Fault detection means to decide whether or not a fault has occurred. This step determines the time at which the system is subjected to the given fault. Fault isolation determines the location of the fault, i.e. in which component a fault has occurred. Fault identification and fault estimation identify the fault and estimate its magnitude.

Besides from detecting sensor faults, the fault diagnosis algorithm has also been used to improve the wave estimation through a sensor fusion quality test. The sensor fusion quality test has been developed in this project to help decide on which three ship responses, e.g. roll, pitch and heave, would be the most suitable for wave spectrum estimation. The sensor fusion quality test should be applied on each combination of three non-faulty signals, and it has been demonstrated that the approach can be used to increase accuracy of sea state estimations considerably. This means that the need for sensors is reduced, thus decreasing the installation costs and at the same time increase the operational value of the onboard decision support systems.

Papers on the topic:

Fault detection for shipboard monitoring and decision support systems

Fault isolation and quality assessment for shipboard monitoring

Fault Detection for Shipboard Monitoring
- Volterra Kernel and Hammerstein Model Approaches

Produced by Adapt A/S