As the global production patterns change gradually into multiple varieties and small batch mode in industry 4.0, automatic production becomes not flexible enough to handle the requirement, therefore the advantages of collaborative robots are becoming increasingly apparent. The safety issues of the collaborative working environment have then also emerged into our sight. This thesis takes collaborative robot as the object of study to discuss a way of improve industrial safety with digital twin application. The purpose of the digital twin application is to achieve the prediction of the working state of the human collaborative robot. The concept of digital twin and the characteristics of human robot collaborative environment is studied, and is compared with other method to specify the target of this thesis.