Fault Diagnosis Techniques in Internal Combustion Engines

Fault diagnosis in internal combustion engines encompasses a suite of methods designed to detect, identify and locate malfunctions that compromise performance, emissions and reliability. Traditional approaches rely on model‐based analysis, comparing measured signals—such as vibration, pressure or rotational speed—against theoretical or empirical models to generate fault indicators. Signal‐based techniques extract features in time, frequency or time–frequency domains using transforms (wavelet, Hilbert–Huang, multisynchrosqueezing) and statistical measures. Recent decades have witnessed the integration of artificial intelligence and machine learning algorithms to enhance diagnostic accuracy and reduce reliance on expert tuning. Techniques range from support vector machines and ensemble classifiers to deep learning architectures, including convolutional neural networks and recurrent networks. Non-invasive strategies employing acoustics, infrared thermography or digital twins minimise interference with engine operation while enabling real-time monitoring. Model-based and data-driven methods are increasingly hybridised to capitalise on the interpretability of physical models and the adaptability of AI, yielding robust systems capable of detecting single and simultaneous faults under diverse operating conditions. These advances bear global significance for reducing greenhouse gas emissions, improving fuel economy and ensuring the reliability of transportation, power generation and industrial machinery.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Fault Diagnosis Techniques in Internal Combustion Engines publication trend

The graph below shows the total number of articles in fault diagnosis techniques in internal combustion engines across all publications each year (not limited to Nature Index journals).

Technical terms

Misfire: A failure of one or more cylinders to ignite the fuel–air mixture, causing power loss and increased emissions.

Time–frequency analysis: A set of signal‐processing methods that decompose signals into joint time and frequency representations to extract transient features.

Digital twin: A virtual model of a physical engine system that mirrors its real-time behaviour for simulation, diagnostics and predictive maintenance.

LSTM RNN: A recurrent neural network architecture with memory cells designed to learn long-range temporal dependencies in sequential data.

Instantaneous angular speed (IAS): The real‐time rotational speed of an engine shaft measured over individual crank angles to reveal combustion irregularities.

References

1. Non-Invasive Techniques for Monitoring and Fault Detection in Internal Combustion Engines: A Systematic Review. Energies (2024).
2. Misfire Detection Using Crank Speed and Long Short-Term Memory Recurrent Neural Network. Energies (2022).
3. Instantaneous Rotational Speed Algorithm for Locating Malfunctions in Marine Diesel Engines. Energies (2020).