A Task Decomposition And Optimization Method For Nuclear Power Plant Maintenance Robot

As an important tool to reduce staff radiation, maintenance robots for nuclear power plant play a significant role in the nuclear power plant accident repairs and routine maintenance and other processes. However, how can a nuclear power plant maintenance robot handle and perform its maintenance tasks to maintenance of nuclear power plants from robot control center? That is the problem of task decomposition and optimization. Task decomposition and optimization is one of the core issues of maintenance robots for nuclear power plants, and is the key to control nuclear power plant to complete maintenance tasks. Therefore, it is need to study the method of nuclear power plant maintenance robots to perform the task decomposition and execute primitive actions after decomposition.Firstly, the method of representation and execution of primitive actions are analyzed in this paper. With a view to particularity of nuclear power plant operation objects, we propose a method for each primitive action taking different definitions and representation, and then using a constrained random generation method and database matching algorithm, the learning algorithm processes for the primitive actions is proposed. The algorithm is self-learning, and has broad applicability and extensibility. By performing the simulation of the learning algorithm and analyzing the simulation results, the correctness and feasibility of proposed algorithms is verified. Secondly, we combine the background knowledge of nuclear power plants maintenance tasks, and use affordance theory to analysis and develop a suitable maintenance problems with nuclear power plant task decomposition methods. This method can be achieved with a combination with primitive actions to achieve nuclear power plant task into the process of primitive actions and to complete the research objectives topics. Thirdly, the Bayesian network model is introduced to decompose the problem of maintenance tasks to build a combining maintenance task decomposition model. Finally, through the design and complete simulation of the model, a detailed analysis of the application of the model, as well as the relationship between the model and the structure of primitive actions are demonstrated in this paper. By analyzing the experimental results, we obtain that the model could make a revise for human factor correction in priori Bayesian network model, and promote the accuracy for actions.