Print Email Facebook Twitter Using a decision tree to analyse results of a simulated execution of operational planning decisions of a container terminal Title Using a decision tree to analyse results of a simulated execution of operational planning decisions of a container terminal Author Van Rhijn, R.A. Contributor Verbraeck, A. (mentor) Lukosch, H.K. (mentor) Li, S. (mentor) Saanen, Y.A. (mentor) Zutt-de Fockert, F. (mentor) Faculty Technology, Policy and Management Department Transport, Infrastructure and Logistics Programme Engineering Date 2015-05-26 Abstract Container terminals are important nodes in the worldwide supply chain. The planning of the daily operations of a container terminal is a complex task and sub optimal decisions can be made due to interconnected sub processes, interrelated decisions, time and uncertainty. If suboptimal planning decisions are made in the planning, consequences will occur during the operation, which cause a decrease in performance. The perfor-mance of a terminal is measured in average Quay Crane (QC) performance. The QC performance determines the turnaround time of vessels. A delay in the turnaround time can cause huge financial loses and a decrease in the reliability of the service a terminal offers. To overcome sub optimal planning decisions, a simulated execution of a planning can be used. If suboptimal planning decisions are made, the consequences will occur during the simulated execution. By analysing the results of the simulation these consequences can be recognised, and thereby the suboptimal decisions can be identified. These decisions should be revised to prevent recurrence of the consequences in the real operation. This can only be done after the planning is finished and before the operation commences, 2-3 hours before the operation. If used not more than 2 Hours are left in the planning process to simulate the planning, analyse the results and revise planning decisions. A simulation takes 15 minutes, including overhead time and performing 3 iterations leaves only around 10 minutes to analyse and 10 minutes to revise decisions. The problem is the lack of a method to use the results of the simulation in such a short time frame. The aim of this thesis is to develop a method that makes it for a planner possible to identify the consequences, define a suitable revision and implement it in the short amount of available time before the operation commences. The research question is stated as follows: How can simulation be used to improve the operational planning of a container terminal before the real execution? The method chosen to solve this problem is a decision tree, which can reveal patterns, identify an assignable cause and define a suitable solution. The developed decision trees are able to reveal the consequences in the results of a simulated execution by checking statistics of the simulated execution with thresholds. A sequence of checks can lead to 3 types of decisions: i. A revision of a planning decision is proposed if the QC performance is lower than desired, the cause could be identified and a solution is available. ii. The tree proposes to lower the planned performance, if the planned performance could not be reached, but no cause was identified or no solution is available. iii. Do nothing is proposed if the desired QC performance is reached, in this case there is no need to make adjustments. To develop this decision tree seven steps are performed; 1. Analysis of the operation and planning, 2. Define the solution space, 3. Perform a root cause analysis, 4. Connect the causes with solutions, 5. Develop the decision trees, 6. Set the thresholds, 7. Evaluate the decision trees. The developed decision tree is evaluated by a case study and expert opinion interviews. In the case study two planning datasets are improved by the use of the decision trees, one reached an increase in QC performance of 1.1 containers/hour and the second of 2 containers/hour. The experts indicated that if the method would be developed further it has the potential to solve the problem. The main conclusions are: - By the analysing the results of a simulated execution of planning decisions with a decision tree, revi-sions of suboptimal planning decisions can be defined - After automation of the decision tree the analysis can be done within 1 minute - The seven steps led to a decision tree that proposed effective solutions during the case study - When the recommendations are processed the decision tree has together with Plan Validation the potential to support a planner to improve their planning and on the longer term planners might be-come more skilled in planning and strategically improvements can be identified From both the case study as from the expert opinion recommendations followed for further research. The main recommendations can be formulated as follows: - The decision trees developed in this research require improvements before implementation is possible - The decision tree should be automated - How the solutions are presented to the planner should be developed (the interface and visualisation) - Plan Validation in combination with the decision tree should be tested in further case studies, by checking it with real planners and by using it on a real terminal Subject container terminalplanningsimulation To reference this document use: http://resolver.tudelft.nl/uuid:d17f95fe-25e1-4d5d-98b0-f03e0bfb50f4 Part of collection Student theses Document type master thesis Rights (c) 2015 Van Rhijn, R.A. Files PDF Thesis_RoosvanRhijn_april2015.pdf 5.51 MB Close viewer /islandora/object/uuid:d17f95fe-25e1-4d5d-98b0-f03e0bfb50f4/datastream/OBJ/view