Abstract: The improvement of ship safety is mainly driven by implementing lessons learnt from maritime casualties into shipping legislation. In this respect maritime casualty investigation and maritime simulation have essential functions which support each other. Casualty reports still have a limited scope and do not provide all necessary data that are needed to get further insight into the accident process. A detailed modelling of the accident process, however, is a key to many safety related tools, such as improved safety training, Decision Support Systems (DSS) or emergency response planning. Simulation can and has done a lot of work to elaborate on some of the existing data gaps with regard to the accident process. Together with other new features, such as the Automatic Identification System (AIS) and the Voyage Data Recorder (VDR) simulation can be used to master new challenges in the maritime safety sector, such as the creation of more realistic training scenarios for Maritime Education and Training (MET) institutions or the design of a new generation of decision DSS. A prerequisite for this step is a wide integration of the different studies and approaches for this problem. This can be achieved by an adequate model of the accident process. This paper is intended to elaborate on the aspects of the relation of maritime casualty investigations and maritime simulation. It is further intended to introduce a model of the maritime accident process and to describe how it can be used to achieve the above mentioned objectives.

　

 Maritime casualty investigation is still a key issue for the development of new and the assessment of existing ship safety regulations. Maritime casualty investigation has to support the following aspects: scientific, legal, educational and enforcement. Especially for the scientific and educational aspect a deep understanding of the accident process is necessary. However the data reality is quite different (Schröder & Zade [16], National Transportation Safety Board [12]). The Human Element (HE) area is insufficiently covered during the casualty investigations. This poses - many questions with regard to the fundamentals and criteria for the assessment of existing and the development of new ship safety regulations.

　

 As a result of the critical review of the deficiencies in the HE area a new approach in marine casualty investigation is required. However, even with a new approach where the HE is put more into the centre of the investigation questions will be left open. This applies to the input of the performance shaping factors (PSF). This is an area where simulation is required and has been proven as valuable in the past (e.g. Howland et al. [6]). In this respect it can be considered that maritime casualty investigation and maritime simulation form a close relation with two important aspects: the investigation aspect and the exploitation aspect.

　

 The investigation aspect describes the part where casualty reports do not answer all questions that researchers or training providers are interested in. This can be either due to missing data or due to behaviour described in the investigation that cannot be explained immediately. On the one hand, simulation can close the data gaps by helping to reconstruct the accident process (Muirhead [11]). On the other hand, it can also help to further elaborate on certain aspects of behaviour. In a realistic scenario operators' behaviour can be studied in an effective way (e.g. Clemmensen [2]).

　

 The exploitation aspect describes the part of the investigation/simulator relation that deals with the lessons learnt from casualties. The primary objective is to avoid a similar accident in the future. There are a variety of tools available to meet this objective. In this paper, the impact on MET is highlighted. However, computer-based decision support systems should also be mentioned, since they can also be developed with the support through simulation.

　

 Both afore mentioned aspects fulfil essential tasks for the improvement of maritime safety. In order to safeguard the effective interaction of both aspects a carefully designed framework is necessary. This can be achieved through a model of the accident process. Such a model can integrate different studies and inputs that come from casualty investigations and simulation. This paper is intended to focus on this framework and to elaborate on the potential of such an integrated approach:

　

 In order to achieve this objective, the paper is structured as follows. In section 2 simulation in MET is reviewed. Section 3 deals with a critical assessment of existing data gaps in the maritime casualty follow-up process. Section 4 gives an overview of the current approaches to overcome the data gap. In section 5 an accident process model is described which could be used for the integration of casualty data and results of the approaches to fill up the missing data. In section 6 benefits of such an integrated approach are highlighted and in section 7 an outlook is given to future research needs in this field.

　

 According to Benedict [1] "simulation is the replication/representation of real processes/models in order to create suitable process characteristics/properties" and can among others be applied to the following purposes:

　

　

 The focus of this paper is the education and training aspect.

　

 Approaches to MET have changed over the last four decades, primarily due to the intensive introduction of simulators in training institutes. The training opportunities offered by simulators have significant advantages compared to the scope of possibilities provided on conventional training vessels. With simulators specific behaviours in certain situations can be taught much better than on "real" ships. This is the main reason, why apart from financial reasons, training ships are less and less used in MET institutions. The 1995 revision of the STCW convention [7] considered the training aspects of simulators and encouraged the use of simulators for MET purposes by providing performance standards in Part B-I/12 guidelines of the STCW Code. Although specific guidelines are not given in Part B-I/12 safety training is mentioned as a method for demonstrating competence in Part A-II of the Code, when among others e.g. training for the response to emergencies and distress signals is described for officers' operational level.

　

 The basis of simulator training is the development of suitable training scenarios. However, according to Muirhead [10] effective simulator training is dependent on:

　

　

 STCW underlines the importance of realistic training scenarios. The development of such realistic training scenarios is demonstrated in Fig. 1.

　

　

 Since realistic scenarios with a close connection to real processes is a primary objective for simulation the question has to be raised: can such realistic scenarios be developed in safety training? This question is relevant when the available data about casualties are considered.

　

 Data from marine casualties are an essential source of scenario development for the training of emergency situations. Simulation is intended to train leadership behaviour and to provide for a better understanding of the whole emergency response process. In order to create realistic simulation scenarios, deficiencies in the emergency response management of previous casualties have to be known, This allows targeting to cover the behavioural reactions that caused major accidents in previous years.

　

 Unfortunately, the data reality does not support the development of such scenarios as described before. An overemphasis of technical details in accident reports causes insightful information on the HE to be missing. This is a major disadvantage when scenarios are to be created that specifically focus on leadership and similar aspects. The data gaps are not only caused by technical overemphasis in the reports. Further underlying factors for the data gaps are: missing uniform approaches to maritime casualty investigation, missing definitions, models of the accident process and its components.

　

 Schröder [14] focused on the existing data gaps in a recent study. For this purpose accident reports of 42 passenger vessel accidents were reviewed. Passenger vessels were used because it was expected that these reports where a high public interest was assumed would provide better information about HE related data. The results were quite negative. Although it is known that accident reports provide only a limited scope of all data collected during the investigation, the quality of HE related data given in the reports cannot be considered as sufficient.

　

 The study showed that particularly the area of the PSF is clearly lacking. As far as the general situation of the crew is concerned Schröder tried to get information that can be grouped as follows:

　

　

 Details of less than 2% of the data categorized of all four data groups above were given in the reports. This is a very remarkable finding, because this means information about such important areas as the status of training of the crew (e.g. which training measures are conducted at which intervals, how many crew members are recent professional entrants etc.) is missing. The same applies to the working conditions (e.g. which watch system is on the ship, how are watches manned, what is the workload of the crew etc.). The list of other missing important points could be extended.

　

 However, it is not only the PSF related data that are missing. In another significant area - the emergency response management - important information is missing, too. Here it is mainly related to decision making problems during the emergency response actions (e.g. handling of the equipment, willingness to cooperate, communication etc.).

　

 The missing HE related data is also confirmed in other studies (e.g. NTSB [12]). However, without this information, it is questionable how the STCW required real-time and realistic simulation of emergency situations can be achieved.

　

 When PSF data are not known important information why and how operators in systems react in the way they have reacted cannot be concluded. The same applies to the emergency response management. Here it is important to know why an accident has developed into a catastrophe and if it could have been prevented through appropriate actions during the emergency response measures. Since these data are underestimated at this time, a new quality of marine casualty investigation has to be developed.

　

 The existing deficiencies in the marine casualty investigation framework could be overcome by a more structured follow-up process to the casualties. A framework consisting of a model of the process/component to be investigated, a related data scheme and a method to get the required data, as suggested in Schröder [14] could lead to improvements in the data situation.

　

 This applies to the PSF data. There are several schemes of PSF data existing in different accident causation schemes or risk assessment methods (a summary can be found in Hollnagel [5])・ The influence of some PSF was targeted in research projects in recent years (e.g. the IMO focused on fatigue) and results to single questions are available, the influence of PSF are quite complex. Therefore a better connection of marine casualty investigation to simulation is required.

　

 Marine casualty investigation is providing the stimulus for further investigations into the circumstances that caused the accident. In an ideal case, it would be possible to collect data and based on a solid data base, areas of concern could be identified. It should be the task of the safety related research to elaborate further on these issues. Different approaches are possible (interdisciplinary studies, comparative studies with other transportation modes etc.). Different tools can be used for these approaches as well. The simulator is a very valuable tool in this respect.

　

 As already mentioned in the introduction, there is a strong relation between the accident investigation and simulation. The investigation aspect of this relation is to understood how the simulator can support the clarification of unanswered questions with regard to human operator behaviour. In a simulated environment the influences and conditions can be controlled and repeated as often as necessary. This helps to explain the data obtained during the casualty investigation (e.g. Fang and Hu [3]). In this respect the reconstruction of the accident in a simulator as described by Muirhead [11] can also contribute to a better understanding of the accident itself and can stimulate measures for safety improvement including training measures on a more objective basis than approaches that depend on expert judgement.

　

 In the previous sections of this paper, the current status of casualty data and their importance for the maritime training process was underlined. It was also outlined in which parts of the accident investigation process simulation can support in order to fill existing data gaps. However, in order to group the results obtained from the different approaches (investigation and simulation) for the reconstruction of the accident process in a meaningful sense a framework is needed. This framework consists of a model of the accident process and a data scheme.

　

5.1 The model of the maritime accident process

　

 The model of the maritime accident process (Schröder [14]) is based on the models of Hahne [4] and van der Schaaf [13] and is shown in Fig. 2. It considers the idea of defences and safety reserve potential in technical systems that should prevent single failures from developing into catastrophes. It also underlines how important it is not only to investigate the accident-causing event but also the initial response measures. This model focuses on the accident process, including the defence mechanisms that have been built into the system when it was designed. By using this model, questions of if and how these defence mechanisms have been proven sufficient can be answered. This is not possible with many other approaches to maritime casualty investigation, such as the IMO approach [8], [9]

　

 The model defines six stages of the maritime accident process:

　

　

 The stages can be described as follows:

　

 The dangerous situation is caused by the interaction of technical, human and organizational failure. It leads to a significant impact on the safe shipboard operational status. This impact is in most cases measurable (e.g. increase of the room temperature). It can, however, also appear as a dimensionless increase of the risk potential to the ship (e.g, due to a hazardous approach of a potential colliding ship before the actual collision). A hazardous situation can only be averted by direct intervention of the operator of the system or due to a defence mechanism that belongs to the safety reserve potential of the system.

　

 The beginning accident is understood as a part-process of the maritime accident process. It describes the phase after the appearance of the dangerous situation till the safety defence mechanisms in the system become effective and/or initial response measures are undertaken. The phase of the beginning accident ends when a planned response to the emergency is necessary after the defence mechanisms alone did not provide the return to the safe shipboard operational status. The phase ends also when, as a result of effective defence mechanisms/initial response measures, the return to the safe shipboard operational status was achieved.

　

 The term near miss describes the result of the phase of the beginning accident when effective defence mechanisms/initial response measures enabled the return to the safe shipboard operational status.

　

 The term accident characterizes the phase of the accident process which is reached after the appearance of the dangerous situation and after the ineffective defence mechanisms/initial response measures did not provide a return to the safe shipboard operational status. In this phase, planned efforts have to be taken in order to respond to the emergency situation. This phase is also characterized by an increase of the danger potential in a way that a total loss cannot be excluded anymore. This is why in this phase, apart from the emergency response measures, preparations for the evacuation and the request of external help is necessary. During the accident phase there is continuous information gathering and situation assessment. The situation assessment is necessary in order to find out if a return to the safe shipboard operational status is still possible or if the ship has to be abandoned. This phase ends either if a return to the safe shipboard operational status was achieved or the ship has to be abandoned.

　

 The term mitigated loss describes the result of the phase accident where as a consequence of effective emergency response measures and/or sufficient safety reserve potential in the system a return to the safe shipboard operational status could be achieved. It also describes the different measures to cancel the evacuation process and to release the external support. In this phase monitoring of the return to the safe shipboard operational status is focused on, no matter how long it takes and no matter if this can be achieved only by passing certain level of return (e.g. if certain systems that were damaged during the accident had to be repaired). Ship abandonment is not characteristic at this phase.