A large number of work safety accidents have exposed that accident has Domino Effect or Avalanche Effect: the chain reaction in system will be triggered when one critical node in system is disturbed, which will finally have an impact on large parts of the system. Exploring the interrelationships of elements on accident chain is very essential for studying the blocked mechanism of the accident. An improved model of dynamic of accident spreading in networks is proposed after analysing Complex Network theory and STDP mechanism. The model has the ability of independent learning of network edge weight, and finally achieves the purpose of strengthening causality and weakening non-causality. We can abstract three kinds of ideal structure networks from a large number of realistic work safety accidents: the random networks, the scale-free networks and small-word networks. This model is simulated and analysed on the three networks, respectively. The process of time dynamic evolution of accident spreading and edge weight is analysed. The simulation results are consistent with the characteristic of actual work safety accident, which indicate that the model can effectively simulate the process of accident spreading dynamics of actual work safety accident and the process of edge weights evolution.