Research on engine power-loss fault diagnosis method based on time-series data mining

Area of Science:

• Automotive Engineering
• Data Mining
• Artificial Intelligence

Background:

• Traditional engine power-loss diagnostics for commercial vehicles are resource-intensive, relying heavily on on-site testing.
• Existing methods face limitations in efficiency and scalability for real-world fleet management.

Purpose of the Study:

• To develop an intelligent, data-driven diagnostic method for commercial vehicle engine power-loss faults.
• To overcome the limitations of traditional on-site testing through remote, online diagnosis.
• To leverage time-series data mining and machine learning for enhanced fault identification.

Main Methods:

• Collected and analyzed real-world operational data from onboard telematics terminals.
• Identified key features correlated with engine power loss: vehicle speed, acceleration, and throttle opening rate.
• Developed a dual-framework strategy: machine learning for 'with driver acceleration intent' data and deep learning for 'without driver acceleration intent' data.

Main Results:

• The proposed intelligent diagnosis method achieved high accuracy and specificity in identifying engine power-loss faults.
• The dual-framework approach effectively distinguished between different fault scenarios based on driver intent.
• Experimental validation confirmed the method's effectiveness and practical potential for remote diagnosis.

Conclusions:

• The study presents a viable pathway for remote, online diagnosis of engine power loss in commercial vehicles.
• The findings lay a foundational framework for the intelligent advancement of vehicle diagnostics.
• This intelligent approach offers a more efficient and resource-conscious alternative to traditional diagnostic methods.