Research on Multi-dimensional Health Evaluation Method for Complex Machinery

This paper focuses on the multi-dimensional health state assessment of complex mechanical equipment, using a fracturing pump as a case study. It proposes a comprehensive assessment framework that integrates the analysis of multiple data dimensions, including real-time operational data, vibration data, and historical maintenance data. A health baseline is constructed by obtaining residual data of the equipment in a healthy state using a Generalized Regression Neural Network (GRNN) and extracting time-frequency domain features. The health assessment model for the vibration data dimension is built by calculating and normalizing the Mahalanobis Distance between the vibration data to be assessed and the health baseline. An equipment operational state evaluation model based on multiple characteristic parameters is established, using these parameters as failure criteria, to quantitatively characterize the Health Index from the real-time operational data dimension. Based on reliability theory, a multi-state transition probability for the main components of the equipment is obtained using a Markov model. The probability is mapped to the Health Index using a weighted average method for defuzzification, thereby constructing a health assessment model for the historical maintenance data dimension. Finally, the health state of the equipment is graded and comprehensively evaluated. This method can provide theoretical support for the preventive maintenance and health management of equipment, while also offering a reference for the intelligent operation and maintenance of complex mechanical equipment.