Abstract: Accurately predicting the jacking force is a crucial for effectively controlling the safety, quality, and schedule of the long-distance pipe jacking project.In this paper, based on the machine learning modeling method of knowledge data fusion, the jacking force calculation physical model was combined with the multi-later perceptron to construct a physics-informed neural network model (PINN) driven by physics and data, the physical mechanism was used to constrain the training mechanism of the neural network, and the improved sparrow search algorithm was introduced to optimize the hyperparameter value of the model.Therefore, the ISSA-PINN jacking force prediction model was established.Taking the water supply project for Zhengzhou-Kaifeng City demonstration area as an example, 524 sets of actual measurement data were selected to validate the model's effectiveness.The results showed that, compared to data-driven models, the ISSA-PINN model achieves high prediction accuracy with 0.07 and 0.17 improvement on test datasets and new datasets respectively.This demonstrated that incorporating physical models have positive effects to reducing over fitting and enhancing generalization of MLP model.The ISSA algorithm demonstrated faster optimization speed and better fitness compared to the standard Sparrow Search Algorithm (SSA) and Particle Swarm Optimization (PSO).The research results can provide reference for jacking force control in pipe jacking engineering construction.