Petri nets for real-time fault diagnosis and recovery in an Intermodal Container Terminal

The growth of intermodal freight transportation, driven by the changing requirements of supply chain, is challenged by two main factors: i) the need to reliably and flexibly respond to changing customer requirements with an effective coordination of equipment and freight flows through various modes, ii) the constraints on the infrastructure capacity, including policy and regulatory issues, as well as a better management of existing infrastructures [1]. In the performance evaluation of the logistics chain as a whole, the efficiency of the intermodal freight terminals is a major goal to pursue. Particularly, the optimization of the activities associated with the intermodal terminal operations and the consequent reduction of operative costs are the most important problems to face for increasing the flexibility and the dynamic capacity of terminals. This means that both material transportation and handling operations have to be analysed; in order to perform a suitable management and well-shaped control policies for freight movements inside the terminal. Then, the control system of a terminal should provide integrated capabilities of fault detection and diagnosis, along with recovery procedures, in order to increase both the system reliability and productivity. Moreover, such a system should be able to respond to the introduction of new services, with a quick service delivery and low service costs. In particular, the control system should i) be provided with automatic system control facilities – service control, operations control; ii) have a hierarchical structure, allowing individual functions to be performed independently, and also allowing the whole system to work in an integrated manner; iii) reduce costs, and shorten service cycle time [2]. As the timeliness of operation executions is a major objective in a terminal, a real-time detection and control system is needed to make the system react promptly to the unpredictable events. Consequently, it is very important to monitor the activities and the resources they use, determine the possible concurrencies by considering which resources are used by more than one activity. Furthermore, the need to respect the scheduled time of operations requires continuous fault monitoring, as well as detection and (on-line) control to provide real-time solutions to possible related problems. In this framework, an intermodal freight terminal in which only containers are handled is considered, with reference to a real case study under concern. The main regulation objective is to minimize the degrade of the system performance when unexpected events occur, and faulty behaviours result from any deviation from the a-priori scheduled daily operating plan. For its structural and operating characteristics, a container terminal can be represented as a Discrete Event System (DES) whose state evolution depends on the occurrence of events such as delays in beginning and/or ending operation execution(s) that can prevent the realization of the a-priori scheduled system functioning. Consequently, Petri nets (PNs) seem to be the most useful formalism able to model the functioning of an intermodal freight terminal as it can be seen in [3]. In this paper, PN will be used to define a hierarchical monitoring structure intended to real-time fault diagnosis and recovery in the considered intermodal container terminal. Because of the complexity of the considered system, a distributed hierarchical discrete event control structure seems to be the best choice. In the proposed control architecture, the lower level is made up of local controllers, each one monitoring and controlling a synchronization point [4], whereas at the higher level a central coordinator acts like a broker, making the local controllers communicate with each other. As a result, the proposed approach consists in designing a decentralized monitoring and control system able to detect, to diagnose, and to correct, through an algorithm of optimization, the consequences of the undesired events. The considered model makes use of Petri nets since they are able to capture the precedence relations and interactions among asynchronous events typical of DES. The detection and control model we propose is based on the PN model of a daily working plan of the container terminal. Furthermore, the control architecture is able to provide information about whenever and wherever an unwanted delay occurs by monitoring the transition firing times, signalling to the regulation module any missed firing, and providing a recovery strategy able to minimize delays and the costs associated with. Such a control structure is being validated based on data about the real case study of the Italian container terminal named Voltri Terminal Europa (VTE), Genova.