Abstract: A state monitoring system based on FPGA and neural network analysis has been proposed to monitor the health status of machine tool spindles. Fault diagnosis is performed when spindle malfunctions are detected, providing a methodology for equipment status monitoring. A rolling bearing fault dataset is used as the input for the bearing's original signals, with computation resource allocation undertaken via FPGA. A deep learning model is established using Python. The developed learning model is described at the RTL level using Verilog language, followed by simulation and synthesis. Finally, simulation verification is carried out on an FPGA development board. A total of 43 066 test sample data were employed, and samples were randomly selected for testing. The system's average accuracy was calculated to be 91.0%, with a relative accuracy reaching 94.5%, verifying the effectiveness of the proposed method and system.