Research On Safety Evaluation Of Inland Unmanned Ships

With the steady advancement of inland navigation development strategy,establishing a smooth,efficient,green and safe inland navigation system is the current development direction of inland navigation,and ensuring the safety of inland navigation is an important step to establish a modern inland navigation system.The rapid development of modern artificial intelligence technology has caused a revolution in driverless technology.Unmanned ships have been more and more widely concerned because of their advantages such as non-fatigue,loyalty to their duties,unaffected by the emotions of the manipulator,all-weather working,and high reliability.The emergence of unmanned ships will bring corresponding changes to inland waterway safety.And it is of great significance to evaluate the navigation safety status of the inland unmanned ships reasonably.Therefore,around this research task,this paper mainly carries out the following research work:（1）Clarifying the indicators selection principles of inland unmanned ships navigation safety status,analyzing the factors affecting inland unmanned ships navigation safety from three aspects of ships,environment and control,and establishing the initial index system for inland unmanned navigation safety.Then formulating the status grading standards for each indicator.The entropy theory and improved correlation coefficient method are used to screen the importance and the global correlation of the initial index system in order to remove indicators that reflect less effective information content and high information overlap of inland unmanned navigation safety status.The concept of the overall identification of the index system is proposed to verify the rationality of the established index system of inland unmanned navigation safety status.（2）According to the index system of inland unmanned ships navigation safety and the state classification standard of each node,the Bayesian network structure of inland unmanned ships navigation safety is determined.The Bayesian parameter estimation method is used to calculate the conditional probability of each node in the Bayesian network through combining with the statistical decision theory and incorporating the prior information of Bayesian network nodes.Then the conditional probability of each node in the Bayesian network is calculated through Matlab programming.Compared with those methods based on expert experience,this method can reduce the subjective uncertainty when calculating the probability of each node in the ship navigation safety Bayesian network and improve the accuracy of conditional probability of each node in the network.（3）With the help of the Ge NIe software platform,the junction tree algorithm is used to establish a Bayesian network evaluation model reflecting inland unmanned ships navigation safety status.The concept of ship navigation safety status in the base period is introduced to determine the status level of each node in the base period.And then calculating the state probability distribution of the ship navigation safety node and its subnodes in the base period.An example is used to calculate the comprehensive probability of the current state of ship navigation safety and the state probability of subnodes of ship factor,environmental factor and control factor.Comparing the change of state probability of each node of current ship navigation safety with that of base period in order to evaluate the navigation safety status of inland unmanned ships in real time.Through analyzing the main reasons that lead to the change of the current ship navigation safety status,providing the decision-making basis for the intelligent collision avoidance of unmanned ships.