The purpose of the work described in this paper is to construct and implement prediction models for optimizing container handling in particular at Cagliari’s Terminal Container. Prediction models are based on heuristic algorithms such as neural networks and classification and regression trees and evolutionary algorithms such as Genetic Algorithms (GA) and Ant Colony Optimization (ACO). These models form part of a Web Oriented Decision Support System, for real time external data acquisition (GPS information, weather information, etc.), providing operators with the information processed in real time. The most commonly used parameter for assessing terminal performance is productivity, namely the number of containers handled in the unit of time considered. This parameter can be associated with the terminal as a whole, or with the ship, the stevedores, each vehicle used, the single operator and related to different time intervals (year, month, week, day, hour and shift). Usually the hourly average is considered for monitoring operations and identifying shortcomings. The rate at which operations are performed can significantly reduce turn round time and thus minimize the loss of productivity associated with the ship’s time in port. Because of the complexity of analyzing decision-making processes two sublevels are defined, that differ for type of decision and time horizon: - The first level, generally organized around a weekly time horizon (from 10 to 1 days prior to ship’s arrival in port), for scheduling operations and activities in the different areas such as, ship, quay, yard, for making decisions that satisfy the different requirements;- The second level, aimed at specific resource allocation (personnel and equipment) on the basis of the decisions made at the first planning level in order to maximise productivity and minimize costs over a time horizon of roughly 24h. Both levels of planning are characterized by temporal fragmentation and uncertain information. The information is received at undefined times and is continually updated, resulting in uncertain content. The strong dependence of the planning process on information flow, means it is necessarily dynamic and makes it difficult to effectively optimize and integrate decisions over sufficiently broad time horizons. The aim of the study is to construct a model for predicting containership arrivals using heuristic-based evolutionary algorithms. The so-called ―Inspect Inspired Algorithms‖ are proving effective tools for solving industrial optimization issues. In this study the different models proposed are implemented in a ―Decision Support System‖, while data are analysed from a temporal aspect adopting a ―learning from data‖ approach. Indeed the observation of real data (actual arrival time of ships and handling in port) form the knowledge base which relies on learning from the past. All discrepancies observed between prediction models and reality, along with other factors governing that condition prediction errors, create a historical base on which models are automatically recalibrated. This approach has the dual purpose of analysing the causes (shortcomings) of prediction errors while refining models for future prediction; an analysis of the causes and effects of recalibrating the models. The proposed DSS can also be used for simulation purposes. In fact the algorithms will be implemented for studying the effects of external variables taken individually or interacting with one another, thus providing a useful planning tool.

Development of prediction models for container traffic

FANCELLO, GIANFRANCO;SERRA, PATRIZIA;FADDA, PAOLO
2010

Abstract

The purpose of the work described in this paper is to construct and implement prediction models for optimizing container handling in particular at Cagliari’s Terminal Container. Prediction models are based on heuristic algorithms such as neural networks and classification and regression trees and evolutionary algorithms such as Genetic Algorithms (GA) and Ant Colony Optimization (ACO). These models form part of a Web Oriented Decision Support System, for real time external data acquisition (GPS information, weather information, etc.), providing operators with the information processed in real time. The most commonly used parameter for assessing terminal performance is productivity, namely the number of containers handled in the unit of time considered. This parameter can be associated with the terminal as a whole, or with the ship, the stevedores, each vehicle used, the single operator and related to different time intervals (year, month, week, day, hour and shift). Usually the hourly average is considered for monitoring operations and identifying shortcomings. The rate at which operations are performed can significantly reduce turn round time and thus minimize the loss of productivity associated with the ship’s time in port. Because of the complexity of analyzing decision-making processes two sublevels are defined, that differ for type of decision and time horizon: - The first level, generally organized around a weekly time horizon (from 10 to 1 days prior to ship’s arrival in port), for scheduling operations and activities in the different areas such as, ship, quay, yard, for making decisions that satisfy the different requirements;- The second level, aimed at specific resource allocation (personnel and equipment) on the basis of the decisions made at the first planning level in order to maximise productivity and minimize costs over a time horizon of roughly 24h. Both levels of planning are characterized by temporal fragmentation and uncertain information. The information is received at undefined times and is continually updated, resulting in uncertain content. The strong dependence of the planning process on information flow, means it is necessarily dynamic and makes it difficult to effectively optimize and integrate decisions over sufficiently broad time horizons. The aim of the study is to construct a model for predicting containership arrivals using heuristic-based evolutionary algorithms. The so-called ―Inspect Inspired Algorithms‖ are proving effective tools for solving industrial optimization issues. In this study the different models proposed are implemented in a ―Decision Support System‖, while data are analysed from a temporal aspect adopting a ―learning from data‖ approach. Indeed the observation of real data (actual arrival time of ships and handling in port) form the knowledge base which relies on learning from the past. All discrepancies observed between prediction models and reality, along with other factors governing that condition prediction errors, create a historical base on which models are automatically recalibrated. This approach has the dual purpose of analysing the causes (shortcomings) of prediction errors while refining models for future prediction; an analysis of the causes and effects of recalibrating the models. The proposed DSS can also be used for simulation purposes. In fact the algorithms will be implemented for studying the effects of external variables taken individually or interacting with one another, thus providing a useful planning tool.
optimizing container handling; containership arrivals predictio; Decision Support System.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11584/72493
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact