To improve accuracy of the Reynolds-averaged Navier-Stokes (RANS) turbulence model simulation, relevant parameters in the standard k-ε model control equation are predicted and modified based on artificial neural networks (ANN). An investigation is analyzed for how the fluid field is affected by the wall injection as part of accuracy improvement processes. After initializing the finite element (FE) model of the fluid field, a standard k-ε turbulence model is used to perform relevant turbulent calculations. Meanwhile, circumstances of how wall injection affected the fluid field in the pipe are also added in the FE model. Transformations of the fluid are analyzed by the FE method. Thus, velocity field distribution after the wall injection is interpreted. Moreover, a neural network(NN) intelligent algorithm is established to predict standard k-ε model parameters. The NN model′s input is considered as a ratio of the velocity component in the streamwise direction and the average velocity of the fluid in the pipe. Parameters in the turbulence model control equation are network′s outputs. Thus, the network predicts standard k-ε model control equation parameters to describe the velocity field trend. The final step is retrieving the outcome parameters into the FEM calculation. Comparing with experiment data, simulation accuracy of the velocity field is significantly improved with the modified turbulence model. It shows that predicting and adjusting the standard k-ε model control equation parameters improves simulation accuracy of the velocity field trends.