Summary: ``Composite services are generally used for complex tasks in service computing. Composite services can be expressed as Bayesian Service Dependence Graph according to data dependencies and control dependencies in the individual services. However, any fault in individual service can result in a failure of the composite service, which makes the discovery of errors in service composite service particularly necessary. Therefore, in this paper, a novel method named `Bayesian Service Dependence Graph for Fault Location' is proposed. Firstly, this method mines data dependence and control dependence in the individual services, and transforms the composite workflow into Bayesian Service Dependence Graph by splitting the output node into data output node and logic output node. Finally, based on Bayesian service dependency graph, Bayesian inference method is used to locate the service errors. At the end of this paper, a simulation experiment is conducted to evaluate the feasibility of the proposed method.''