Abstract: The stress state is the fundamental for evaluating the soil strength and stability，playing a crucial role. However, during the stress testing, local damage and other uncertain factors may lead to partial sensor data missing, causing the existing three-dimensional stress calculation method to fail. To accurately restore the soil stress state during data missing, a three-dimensional stress calculation method was developed based on three-dimensional stress testing principles, incorporating axisymmetric and one-dimensional compression characteristics. The three-dimensional stress, principal stress, the first invariant of stress I₁, the second in variant of stress J₂ and stress Lode angle of a sandy soil foundation under one-dimensional compression conditions with different data missing were calculated and compared to results with complete data. The results show that the method is highly accurate; as the load increases, the relative error decreases and converges. The principal stresses, the first invariant of stress I₁, the second invariant of stress J₂ and the stress Lode angle align with one-dimensional compression response, suggesting that this calculation method supports advanced data mining. This study offers a novel approach and a practical method for fully utilizing the test data.

Key words: three-dimensional stress testing, sensor data missing, transformation matrix, principal stress