People Education Research Industrial Agenda  
Current Research Archive Publications PhD theses Software      


Artificial intelligence for the control of a hopper dredger

Project members:  J. Braaksma, prof. R. Babuška, J.B. Klaassens, C. de Keizer
Keywords:  Adaptive and learning control, Optimal and model predictive control, System identification and estimation, Industrial processes
Sponsored by:  IHC Systems, Senter
This project is a cooperation of the Delft Center for Systems and Control with IHC Systems, a company specialized in the development and manufacturing of automation systems for dredgers. Although modern trailing suction dredgers are equipped with advanced dynamic positioning and tracking systems, there is need for an on-board decision-support system that will advise the operators on a control strategy leading to optimal dredger performance under given operating conditions.

Figure: Artist's impression trailing suction hopper dredger.

The dredging process can be subdivided into two main subprocesses: trailing (propulsion of the ship) and dredging (excavation of the soil from the sea bed and its transport to the ship). Set-points for manipulated variables influencing these processes are determined by two human operators. Consequently, the performance and efficiency of the entire dredging process heavily depend on the experience and insight of these operators.

Changes of external variables that have large influence on dredging efficiency, such as the type of soil, dredging depth, water current, etc., require that the operators must constantly change the important settings of the manipulated variables. These include the propeller pitch, the pump drives, the visor angle, swell compensators, etc., when these actuators are controlled manually, or the corresponding set-points for trail speed, mixture speed, soil-water mixture density, etc., when controlled automatically.

An important constraint is the limited amount of energy available on-board. Proper distribution of the energy among the different subprocesses is thus crucial. In addition, different operating strategies can be used in different dredging projects, such as the maximization of the production rate vs. optimization of efficiency and awareness of maintenance and fuel costs.

The goal of this project is to develop an adaptive decision-support system that will advises the operators on the most suitable control strategy, given a specified goal, such as the minimization of the integral dredging costs per m3 or the maximization of the production per time unit. The system will make use of available knowledge in the form of (partial) mathematical models of the process and will also involve on-line learning and adaptation during operation.


Back to top

Last modified: 30 November 2009, 9:52 UTC
Search   Site map