Currently probability and extent of the loss of especially serious accidents is still relatively high. According to SAWS statistics, major accidents occurring every year in the two cases above between 2013 and 2015. Due to the risk of major accidents in the high degree of concentration of matter and energy, the accident led to huge direct and indirect economic losses which the amount is often shocking. On account of the limited traditional insurance market, accident compensation cannot meet the requirements for damages particularly significant risk. Third Plenary Session of the party's eighteen requirements to promote national governance systems and governance modernization. If the management for the major accident risk could be promoted to the governance level, it is bound to favor the accident loss evaluation system, integrated collaborative governance, deepening the crisis management and the maintenance of social stability. In this paper, the risk of major accidents assesses by a Bayesian network model. Bayesian network is a common model of the system feasibility. Bayesian networks require detailed conditions given in transition probability between each node. We draw the entire system before the accident as a system. The occurrence of major accidents proof that the system reliability comes down. Through the establishment of a hierarchical Bayesian network model, we solve the particular risk of major accidents system reliability function and to derive the critical value of the crisis. And on this basis, considering repair of the system, we collect data censored by the timing and the number scheduled, to take strategy of "repairing the old as old" and "repair the old as new", improving system reliability. At the same time, we select the administrative, equipment, education as externally observable characteristics of key data indicators, establishing degradation model to the performance of "soft accident." When an accident occurs, the system is still working. But system performance cannot achieve reliability requirements. When computing the probability of "soft accident" occurred, we highlight the impact of the human factor. This article focuses on the perpetrator's liability risks. Our responsibility is to deal with the risk of civil liability of legal institutional arrangements. Civil liability is the legal responsibility of a compensatory nature and the compensation liability is reflected in the property, pursuing a "no harm, no compensation, which shall not be more than harm" principle. When we argue that the establishment of public liability insurance helps to defuse the risk of major accidents, the insurance companies underwriting the insured engaged in production or other activities in a fixed place or location, due to an accident caused when the law should bear the damage to property or injuries to others liability. Finally, we propose the regulation sound legal system for a basis of liability insurance, the establishment of major accident public liability insurance to rich insurance products, periodic quantitative detection of "soft accident" to timely warning the responsible party.
