Abstract : Starting from the behavior of complex systems, this paper studies the feasibility of state prediction for complex mechanical systems, pointing out that short-term prediction of complex mechanical systems is feasible, and proposes a chaotic-based approach to state prediction for complex mechanical systems. Based on the phase space construction method of various signal data sequences of the mechanical system state, a mathematical model for determining the maximum prediction time of the complex system state is studied. Keywords : Mechanical system; Prediction; Chaos 0. Introduction The fault diagnosis of complex mechanical systems that we usually study treats the state of the mechanical system at a certain time during its use as a static point, studying its state properties and characteristics. However, if we expand the research scale and study the entire use process, we will find that the state continuously evolves and changes with the different periods of use, which is a dynamic process. Therefore, what we study is the motion trajectory of the mechanical system's state, that is, the state history of the mechanical system. Prediction of complex mechanical systems is to predict the motion state of the trajectory after a certain time based on the known motion trajectory of the mechanical system before a certain time. At the same time, the temporal relationship between state points is taken into account. In terms of scale, it has evolved from studying attractors as a static set of points to studying them as dynamic motion trajectories. A large number of chaos research results have proved that the basic characteristics of chaos determine that the prediction of chaotic motion has a maximum predictable scale. Within the predictable scale, the trajectory of motion can be predicted [1]. The emergence of chaos and fractal dimension theory allows us to calculate the fractal dimension of a seemingly chaotic time series and characterize its structure. At the same time, we can also obtain information about the predicted value from a time series. Therefore, in this paper, we use Kolmogorov entropy theory to study the maximum predictable time of the system state. 1. Establishing a mathematical model of the maximum predictable time Mathematical model of the maximum predictable time of the mechanical system state (full text download)