Applying voiceprint technology to the marine field, its robustness is only important, this paper establishes a voiceprint recognition model based on TDNN-Transformer, which improves its robustness compared with the traditional voiceprint recognition model, and adapts more to the real cabin environment. The highest correct result of 98.43 was achieved on the zhvoice speech dataset, and the stable correct rate reached about 95%. A virtual test platform for voice interaction was written, which was interconnected with the turbine simulator via UDP protocol to realize the virtual testing of control commands, and the experimental results show that voiceprint recognition has enough in robustness to be applied to the field of ship turbines in the virtual test environment.