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

TL;DR

This paper introduces a new intelligent method for diagnosing engine power-loss faults in commercial vehicles using time-series data mining, reducing the need for on-site testing.

Contribution

A dual-framework diagnostic strategy combining machine learning and deep learning for remote engine fault detection.

Findings

Key features like vehicle speed and throttle opening rate were identified as strongly correlated with engine power loss.

The proposed method achieved high accuracy and specificity in fault identification.

The approach enables remote, online diagnosis of engine power-loss faults in commercial vehicles.

Abstract

Traditional diagnostic approaches for engine power-loss faults in commercial vehicles are limited by their heavy reliance on on-site road testing and high consumption of human and material resources. To address these limitations, this study proposed a new intelligent diagnosis method based on time-series data mining. Analyzing real-world operational data collected from onboard telematics terminals, identified key features strongly correlated with engine power loss, including vehicle speed, acceleration, and the rate of change of throttle opening. Building upon these features, a dual-framework diagnostic strategy was developed: the data were first categorized into two groups, “with driver acceleration intent” and “without driver acceleration intent”, based on the rate of change of throttle opening. For samples with acceleration intent, multiple machine learning algorithms were employed to model and diagnose vehicle power performance; for those without driver acceleration intent, a deep learning model integrated with classification techniques was introduced to detect latent power-loss faults. Experimental results demonstrated that the proposed method achieved high accuracy and specificity in fault identification, confirming its effectiveness and practical potential. This study provides a viable pathway toward remote, online diagnosis of engine power loss in commercial vehicles and lays a foundational framework for the intelligent advancement of this field.

Figures