Study On Risk Analysis And Early Warning In Ship Maintenance System

Currently risk warning study, which takes account of the safty issue, is a popular research topic in the field of ship management. This dissertation takes ship maintenance system as the research object. The aim of this research is to establish a theoretical framework for risk analysis of ship engine system. Different studies include risk source identification, system diagnosis, risk prediction strategy, and the construction of the risk warning system, etc. Finally, the effectiveness of the strategy method is verified by a real case, which shows a good agreement. The study, including the different identification strategies and methods, has an important theory significance and also the prospect of future application in the industry of ship management.(1) The research introduces the concepts of maintenance risk and maintenance accident in the study of risk management. In accordance with the relevant industry standards and the idea of this study, risks are classified into different groups and being compared with each other. The relationships among the driver, carrier, environment and management factor are setup. Through those relationships we can easily find where and how the dangerous sources happen. The above analysis is combined with the equipment management experiences of senior seamen and experts. In this way, we establishe a basic structure for early-warning risk management system.2) Dangerous source identification is the initial issue of maintenance risk early warning, and we established a knowledge engineering system to solve the problem. The idea of knowledge based engineering ssytem is based on the experiences database. As a complete study of the identification of dangerous source, we introduced the application method of the knowledge employing, knowledge expression, knowledge acquisition, etc. Experiential knowledge of the identification of dangerous source is firstly obtained; then as the knowledge acquisition method, we use decision tree which is based on information entropy theory; after acquisation of condition identification knowledge with monitor instances, a so-called C4.5algorithm was used to measure the importance of various monitoring attributes for finally realising the condition identification. Ship equipment faults sample table is established and examined. Obtained identification results are shown to be quite reasonable and reliable.(3) Clustering analysis in the field of statistics were employed in this study, since risk potentials tend to change frequently and being in an unstable state. Take the evaluation method of Safety Management System (SMS) as an example. Firstly, K-means clustering method and Hierarchical clustering method were use to divide the monitoring data into several typical categories. Secondly, set pair analysis is used for state identification from being 'same','different', and'opposite'. The three aspects can be used to calculate and determine the relation degree with the typical states of SMS. In this way we obtain the certainty and uncertainty of SMS running state. This method may carry out quantitative and qualitative information together for identification.(4) Risks of a single equipment and the overall system (i.e.:the ship engine) are treated and analysed separately in this paper. A predicting model is constructed according to the investigation of the current situations of shipping, the research and analysis of the international maritime pacts, codes, the requirements of Port State Control (PSC). The model was established using BP neural networks method to identify the various marine equipment emergency ultimately.(5) According to the theory of early warning for the ship maintenance, the forewarning management system is setup. We investigated the management mechanism, data structure, the form of software, early warning signals and the output mode, etc. With the help of Access, Excel and VB6.0, the software of forewarning management system is programmed for endusers. In this way, this dissertation tries to offer an effective and useful method to forewarning maintenance risk or reducing the loss of maintenance risk caused by human misjudging.