Ion Current-Based Knock Detection using Convolutional Neural Networks
Björnsson, Ola LU
and Tunestål, Per
LU
(2024)
7th IFAC Conference on Engine and Powertrain Control, Simulation and Modeling, E-COSM 2024
58.
p.261-266
- Abstract
Engine knock, an abnormal combustion phenomenon, can lead to significant engine damage and reduced performance in spark-ignited internal combustion engines. Traditional knock detection methods using acoustic sensors are prone to errors and limited by their sensitivity to sensor placement and engine noise. This study employs a Convolutional Neural Network (CNN) to classify knocking events using ion current measurements from a 13-liter compressed natural gas (CNG) heavy-duty spark-ignited engine. The dataset comprises ion current and in-cylinder pressure measurements, with the ion current serving as the CNN input and the pressure trace used to calculate Maximum Amplitude Pressure Oscillations (MAPO) for labeling knock intensity. The data... (More)
Engine knock, an abnormal combustion phenomenon, can lead to significant engine damage and reduced performance in spark-ignited internal combustion engines. Traditional knock detection methods using acoustic sensors are prone to errors and limited by their sensitivity to sensor placement and engine noise. This study employs a Convolutional Neural Network (CNN) to classify knocking events using ion current measurements from a 13-liter compressed natural gas (CNG) heavy-duty spark-ignited engine. The dataset comprises ion current and in-cylinder pressure measurements, with the ion current serving as the CNN input and the pressure trace used to calculate Maximum Amplitude Pressure Oscillations (MAPO) for labeling knock intensity. The data was balanced and split into training, validation, and test sets to ensure robust performance evaluation. The CNN model achieved a 74.5% accuracy on the test set, with high precision in distinguishing between no-knock and heavy-knock cycles but more difficulty with mid-knock events. ROC curve analysis using the one-vs-rest method was performed, showing Area Under the Curve (AUC) values of 0.96, 0.83, and 0.92 for no-knock, mid-knock, and heavy-knock classifications, respectively.
(Less)
- author
- Björnsson, Ola
LU
and Tunestål, Per
LU
- organization
- publishing date
- 2024-11
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- convolutional neural network, engine control, ion current, Knock detection, machine learning
- host publication
- IFAC-PapersOnLine
- volume
- 58
- pages
- 6 pages
- publisher
- Elsevier
- conference name
- 7th IFAC Conference on Engine and Powertrain Control, Simulation and Modeling, E-COSM 2024
- conference location
- Dalian, China
- conference dates
- 2024-10-30 - 2024-11-01
- external identifiers
-
- scopus:85214198432
- DOI
- 10.1016/j.ifacol.2024.11.154
- language
- English
- LU publication?
- yes
- id
- b4dcd3b0-aad7-40d8-809e-ead2538c4fbf
- date added to LUP
- 2025-02-25 12:23:27
- date last changed
- 2025-04-04 14:45:09
@inproceedings{b4dcd3b0-aad7-40d8-809e-ead2538c4fbf,
abstract = {{<p>Engine knock, an abnormal combustion phenomenon, can lead to significant engine damage and reduced performance in spark-ignited internal combustion engines. Traditional knock detection methods using acoustic sensors are prone to errors and limited by their sensitivity to sensor placement and engine noise. This study employs a Convolutional Neural Network (CNN) to classify knocking events using ion current measurements from a 13-liter compressed natural gas (CNG) heavy-duty spark-ignited engine. The dataset comprises ion current and in-cylinder pressure measurements, with the ion current serving as the CNN input and the pressure trace used to calculate Maximum Amplitude Pressure Oscillations (MAPO) for labeling knock intensity. The data was balanced and split into training, validation, and test sets to ensure robust performance evaluation. The CNN model achieved a 74.5% accuracy on the test set, with high precision in distinguishing between no-knock and heavy-knock cycles but more difficulty with mid-knock events. ROC curve analysis using the one-vs-rest method was performed, showing Area Under the Curve (AUC) values of 0.96, 0.83, and 0.92 for no-knock, mid-knock, and heavy-knock classifications, respectively.</p>}},
author = {{Björnsson, Ola and Tunestål, Per}},
booktitle = {{IFAC-PapersOnLine}},
keywords = {{convolutional neural network; engine control; ion current; Knock detection; machine learning}},
language = {{eng}},
pages = {{261--266}},
publisher = {{Elsevier}},
title = {{Ion Current-Based Knock Detection using Convolutional Neural Networks}},
url = {{http://dx.doi.org/10.1016/j.ifacol.2024.11.154}},
doi = {{10.1016/j.ifacol.2024.11.154}},
volume = {{58}},
year = {{2024}},
}