LEE- Deniz Ulaştırma Mühendisliği Lisansüstü Programı
Bu topluluk için Kalıcı Uri
Gözat
Konu "Electrophysiological data acquisition" ile LEE- Deniz Ulaştırma Mühendisliği Lisansüstü Programı'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeDesign of seafarer-centric safety system; mental workload (MWL) prediction(Lisansüstü Eğitim Enstitüsü, 2022) Özsever, Barış ; Tavacıoğlu, Leyla ; 711394 ; Deniz Ulaştırma MühendisliğiIt is known that human factor has a major effect on maritime casualties that cause great harm to environment, economy and maritime sector. It was stated that while human error is the primary contributor of accidents, a good part of collisions and groundings were related to mental workload (MWL) of watchkeeping officers. Automation, mechanization and the introduction of new technologies had changed the working conditions together with reducing the number of crew and increasing the MWL of operators. This clearly indicates that human element related issues will continue to be one of the major issues in marine transportation assets. In maritime-related studies, it has been analysed mostly how the ship's environment, working period and other factors affect the seafarers. Almost all maritime-related studies couldn't have a potential to develop MWL prediction system for maritime operations aspect. However, lots of studies on drivers and pilots, have produced successful results for MWL prediction. Taking into consideration the fact that MWL has major contribution to maritime casualties, the development of real-time MWL prediction system is vitally essential for ships. By implementing the similar measurement techniques used in the studies on drivers and pilots, to maritime transportation, this study aims to classify the physiological responses of the operators that can produce an output for state of officer on duty as "Safe" or "Risky" from the collected physiological data and task load data during the seaborn operations. This study predicates on the theories which are the statement "minimum performance requires sufficient behavioural activity" of Sheridan and Simpson (1979) together with inverted U function of Yerkes and Dodson (1908) which presents the relationship between arousal and performance. Moreover, the theory of Young et al. (2015) which presents the relationship among mental workload, performance, task demand and resource supply and indicates the overload region, guides this study in terms of building the structure of the experimental research. By being predicated on the above-mentioned theories, this study aimed to design Cognitive Seafarer - Ship Interface (CSSI) which is a main part of Seafarer-Centric Safety System. The physiological data of the 17 junior deck officers (12 subjects performed navigation scenario, 5 subjects performed cargo operation scenario) was recorded according to the design. By being correlated with the performance of the officer, the change of physiological responses of the subjects were analysed in low and high task load levels. The medical decision-making process, which deduced "Safe" or "Risky", was run for this change. For performance measurement that is a part of triangulated measurement strategy (Wierwille and Eggemeier, 1993), Officer Performance Model which is used for MWL classification, was developed for navigation and cargo operation tasks. Additionally, the inputs of Task Load Estimator were defined as data transcription from navigational aids according to results of classification. In summary, the following process were done and results were found. Firstly, the navigation and cargo operation scenarios were created to simulate ship environment. The difficulty level of navigation scenario was gradually adjusted (in order to prevent acquired skill) according to traffic density, visibility and geography by combining in 4 steps. The difficulty level of cargo operation scenario was gradually adjusted according to type and number of operation and operation period corresponding to a real cargo operation by combining in 3 steps. Task load assessments of the scenarios were carried out according to Operator Function Model (OFM-COG) and its sample implications in literature. The results of NASA-TLX scores of the subjects supported the increase of task load levels of the scenarios. ANOVA results showed that there are significant differences in the NASA-TLX scores of 5 different dimensions and in total, among 4 steps which have different task load levels for navigation scenario. Similarly, ANOVA results showed that there are significant differences in the NASA-TLX scores of 3 different dimensions and in total among 3 steps which have different task load levels for cargo operation scenario. According to the subjective assessments of the subjects, MWL increased during the both of navigation and cargo operation scenarios. Secondly, ROC curve analysis was performed for validation of developed officer performance model. Recorded performances of the participants were evaluated as "safe" and "risky" for each task by one ocean going Master expert for navigation tasks and by one ocean going Chief Officer for cargo operation tasks. According to the ROC curve analysis, developed officer performance model was validated with high significance and AUC values. These results showed that the developed officer performance model can be used in any study focused on performance measurement in navigation and chemical tanker cargo operations. Being validated measurement method, performances of the subjects showed that there is a negative significant correlation between performance score and task load in both of navigation and cargo operation tasks. With the distinction of the task load as high task load and low task load, the performance scores were also found significantly different in low and high task loads for both of navigation and cargo operation tasks. Thirdly, physiological responses of the subjects were often differentiated between low and high task loads. Although the change of time-based heart rate variability (HRV) features was not found meaningful according to literature during the increase of task load, the change of frequency-based, time-frequency and nonlinear HRV features were found significant and meaningful during the increase of task load. Moreover, the change of some electrodermal activity (EDA) features and some eye responses were found significant in this study. However, the change of EDA responses was not found strongly correlated with the increase of task load. This can be explained by the fact that electrodermal activity occurs in stressful conditions rather than mental workload. The "frustration" scores of the NASA-TLX supported the fact that the subjects didn't feel so stressed during the tasks. On the other hand, the change of pupil diameter features was found significant and meaningful during the increase of task load in navigation tasks but in cargo operation tasks. Additionally, the change of blink frequency features varied across the scenarios. The variable results of eye responses are thought that the selectivity of eye blinks and pupil diameter to MWL is low according to literature. Additionally, the reason of the fact that the change of some eye features was significant during the increase of task load is thought to be related with the characteristics of eye responses that pupil diameter change is correlated highly with error rate and blink rate increases in incorrect responses rather than correct responses. Therefore, these significances can be explained with the decrease of performance together arising from the increase of task load. On the other hand, the correlations between HRV and EDA features, HRV and eye features, EDA and eye features were found significant and meaningful in mental workload theory. Classification process was carried out with artificial neural network (ANN) code and "Classification Learner" tool of Matlab 2020a. Although the results of the classifications of the subjects' physiological responses on high and low task loads in this study did not give very good accuracies, compared with the studies in literature, they gave sufficient results. The classification accuracies, 75.7% in testing, 83.3% in all for navigation tasks, 80.0% in testing, 92.5% in all for cargo operation tasks and 61.3% in testing, 77.0% in all for cross-task classification have been found similar to those stated in the related studies whose mental workload and stress classification accuracies vary between 70.48% and 98%. According to classification efforts of physiological responses on high task load and low task load levels and performance scores of the subjects, the red lines of task demand became appear in this study. Continuing from the aim of Orlandi and Brooks (2018) and the contributions to MWL prediction in marine engine operations of Yan et al. (2019), the red lines of task demand in ship navigation was tried to determine in this study. Classification of physiological responses and the distinction of the task loads according to the performances of the subjects have ensured the task load to be separated as high task load and low task load. Thus, the inputs of the Cognitive Seafarer-Ship Interface (CSSI) were formed with the outputs of high task load details for navigation and the physiological responses given as features (classified in this study). CSSI processes the task loading together with physiological data of the officer and gives an output as "Risky" for safety of navigation in "The future Seafarer-Centric Safety System design" to be used on ships or at the Shore Control Centre for autonomous ships in future. Consequently, this study will contribute to literature, being the first study in terms of predicting MWL for navigation and cargo operations in maritime transportation. In addition, this study will be a guide for future studies as it reveals the design of the "Seafarer-Centric Safety System" to be developed in order to minimize maritime casualties.