Sensing data from vessel operations are of great importance in reflecting operational performance and facilitating proper decision-making. In this paper, statistical analyses of vessel operational data are first conducted to compare manual noon reports and autolog data from sensors. Then, new indicators to identify data aberrations are proposed, which are the errors between the reported values from operational data and the expected values of different parameters based on baseline models and relevant sailing conditions. A method to detect aberrations based on the new indicators in terms of the reported power is then investigated, as there are two independent measured power values. In this method, a sliding window that moves forward along time is implemented, and the coefficient of variation (CV) is calculated for comparison. Case studies are carried out to detect aberrations in autolog and noon data from a commercial vessel using the new indicator. An analysis to explore the source of the deviation is also conducted, aiming to find the most reliable value in operations. The method is shown to be effective for practical use in detecting aberrations, having been initially tested on both autolog and noon report from four different commercial vessels in 14 vessel years. Approximately one triggered period per vessel per year with a conclusive deviation source is diagnosed by the proposed method. The investigation of this research will facilitate a better evaluation of operational performance, which is beneficial to both the vessel operators and crew.
With the current revolution in Shipping 4.0, a tremendous amount of data is accumulated during vessel operations.
Data quality (DQ) is becoming more and more important for the further digitalization and effective decision-making
in shipping industry. In this study, a practical DQ assessment method for raw data in vessel operations is proposed.
In this method, specific data categories and data dimensions are developed based on engineering practice and existing
literature. Concrete validation rules are then formed, which can be used to properly divide raw datasets. Afterwards,
a scoring method is used for the assessment of the data quality. Three levels, namely good, warning and alarm,
are adopted to reflect the final data quality. The root causes of bad data quality could be revealed once the internal
dependency among rules has been built, which will facilitate the further improvement of DQ in practice. A case study
based on the datasets from a Danish shipping company is conducted, where the DQ variation is monitored, assessed
and compared. The results indicate that the proposed method is effective to help shipping industry improve the quality
of raw data in practice. This innovation research can facilitate shipping industry to set a solid foundation at the early
stage of their digitalization journeys.