Session: 04-01: Powertrain Diagnostics and Control Strategies

Paper Number: 140748

140748 - Virtual Pressure Sensor Based on Ion Current Measurements Using Artificial Neural Networks 

Abstract: 

In-cylinder pressure measurements serve as the gold standard for combustion diagnostics in Internal Combustion Engines. However, the associated costs render it unfeasible to equip engines with in-cylinder pressure sensors.  For spark-ignited engines, using the more affordable in-cylinder ion current measurements as an alternative to the traditional pressure sensors has shown promising results for combustion diagnostics but has not been widely adopted, most likely due to concerns regarding the robustness.

The recent advancements in machine learning and big data have attracted significant attention.
Given the strong correlation between ion current measurements and in-cylinder pressure, combined with the simplicity of collecting a large amount of data, ion current measurements are a good candidate for building machine learning models.

In this paper, we show the feasibility of utilizing Artificial Neural Networks [ANN] to estimate the in-cylinder pressure using the ion current measurement as an input, effectively creating a virtual pressure sensor. Additionally, we calculate different combustion metrics that can be useful for combustion diagnostics and compare them with the metrics calculated on the measured in-cylinder pressure trace. Some of the combustion metrics are the following: Indicated Mean Effective Pressure [IMEP], Peak Pressure Location [PPL], maximum pressure, and heat release-based measurements such as total heat release and CA50. Furthermore, we explore the possibility of estimating the combustion metrics directly from the ion current measurements also using ANN, and compare their performance in relation to the metrics calculated from the virtual pressure sensor.

Presenting Author: Ola Björnsson Department of Energy Sciences

Presenting Author Biography: I am a Ph.D. student in the Department of Energy Sciences at Lund University's Faculty of Engineering, where my research focuses on combustion diagnostics and control in spark-ignited engines through ion current measurements. My work aims at developing and refining combustion diagnostic and knock detection methods in order to enhance engine efficiency.
Prior to pursuing my Ph.D., I completed my Master's degree in Industrial Engineering and Management, specializing in finance and risk, in 2019. During my master's studies, I developed a keen interest in statistical signal processing and machine learning, which I continue to apply in my current research to extract meaningful insights from combustion data.

Authors:

Ola Björnsson Department of Energy Sciences
Per Tunestål Department Of Energy Science