The Research Of Operator Behavior Reliability For Emergency Tasks In Digitalization System Based On Neural Network

With the extensive application of digital system, the reliability of system equipment (hardware and software) has been increased, the operating environment has also been improved significantly, but as an extremely important part in the complexity human - machine system, people's role in the system is more important and prominent. Especially when accident has happened, manipulator's emergent actions directly related to the final disposal of the accident and system's security. Therefore, the reliability of the manipulator's emergent actions plays a vital role to the safe operation of complex industrial systems, and it is also very important to reduce secondary accident and avoid the hurt to operators.The study on the reliability of the manipulator's emergent actions,is to explore the connection between the factors for the reliability of the manipulator's emergent actions and the level of the manipulator's emergent actions, try to predict the level of the reliability of the manipulator's emergent actions under the main factors for the reliability of the manipulator's emergent actions, and hope provides one method for the industrial system, which make the level of the manipulator's emergent actions highest through adjusts each influencing factor. But there are many factors for the reliability of the manipulator's emergent actions, and these factors are not the same degree of influence . Therefore, assessment is difficult to use a mathematical model to express, it is a nonlinear problem.Artificial neural network is widely used in all kinds of problems, due to its ability of inherent non-linear processing,adaptive learning and high fault-tolerance. The Back Propagation Neural Network is the one of many types, it is one style of Artificial Neural Networks, which is a multilayer feed forward network,having strong nonlinear mapping ability. We apply the LM algorithm in network training, improve greatly the convergence speed of neural networks and the prediction accuracy. Considering the complexity and nonlinear to research the reliability of emergency behavior, this paper adopts BP Neural Networks to modeling and analysis, the primary research is as follows:1. To analyze the basic characters of human &system in complex industrial system. Focus on two points: one is the study to the basic characters of human &system; another is the research to the emergency of the human & system.2. To research the manipulator's emergent actions fault mechanism in digital system,to analyze comprehensively all the human error factors to cause human error in digital system,to analyze importantly the human error factors of emergency behavior on man - machine interaction(the main factors for the reliability of emergent actions).3. To make a system research on the constructing and training of BP network model. To analyze Problems exist in BP learning algorithm,Statement corrective method of Levenberg-Marquard BP, and has discussed its characteristic.4. Design for neural network model of Manipulator's emergent actions in digital system, including fix the network structure, learning parameters and learning algorithm.5. The system carries on the confirmation in simulation software MATLAB, Through the experiment data,it indicates using BPNN in my task is feasible, and satisfy the request of precision,it really prompts a convenience tools for predicting the reliability level of manipulator's emergent actions.The thesis finally summarized the research work of this article,and discussed some solution of questions well as the further research work prospect.