This study is related to damage identification method in regression analysis, we evaluate probability of failure for the identification. Deduced from the true incidence of damage to the diagnostic results of Bayesian estimation, evaluation of probability of failure. In addition, the improvement in accuracy by weighted regression, we focus on the influence of parameters is given to the failure probability weighted regression. Weighted regression affects the accuracy of the estimated size of damage. Case of underestimation, An accident occur due to missed damage, in the case of overvaluation, excessive repair inspection will be generated, and these are compared with the same one-dimensional two in comparable economic impact and social impact damage such as human difficult. Therefore, risk assessment for the evaluation of these effects. It is considered the same event can be compared with each one-dimensional. Thus, the efficient operation of the system is possible. Content is about impact assessment study of the distribution of weights for high accuracy evaluation of probability of failure, the reduction of risk.