Design of a Diesel Engine Fault Diagnosis System Based on Performance Indicators

This paper designs a fault diagnosis system based on subspace identification method, which can determine whether the key indicators of the measurement points collected by the ship system have faults, isolate the specific sensors that have faults, and monitor and alarm in real time. The main task of data-driven process monitoring methods is to quickly and accurately detect faults that occur in industrial processes, and to identify, isolate, and recover faults to ensure the stable operation of the system. However, recent research results and feedback from industrial practice indicate that not all process failures will affect the final product quality of the system. On the contrary, ignoring the alarm of such failures can significantly reduce unnecessary downtime and maintenance time, thereby greatly improving the production efficiency of the system and reducing maintenance costs. The development of subspace identification methods has made it possible to directly identify the kernel space of dynamic systems through data, enabling fast and effective modeling of steady-state data for fault diagnosis system design. In view of this, this article adopts data-driven methods such as multivariate statistical analysis and subspace identification to conduct in-depth research on KPI based fault detection methods.