This early ignition, often caused by high surface temperatures, residual gases, or oil droplets, can lead to engine knocking, mechanical wear, and in some cases, complete engine failure. Because the factors behind it are varied and hard to track, traditional methods have struggled to identify pre-ignition reliably, let alone in real time.
Many of the same reasons that hydrogen is such an attractive, clean alternative to traditional fuels make it more prone to pre-ignition. Hydrogen is more flammable and can ignite very easily.
Dr. Abdullah U. Bajwa, Research Engineer in SwRI's Powertrain Engineering Division
The Current Study
Scientists from Southwest Research Institute (SwRI) and the University of Texas at San Antonio (UT San Antonio) are collaborating to develop a method that will address this problem using pressure sensors and machine learning. The real-time, pre-ignition sensing will center around the use of laboratory-grade pressure sensors embedded within the engine to collect fine-grained data on cylinder pressure throughout each combustion cycle.
They hope that these pressure readings reveal combustion behaviour in high detail, enabling them to distinguish between normal and abnormal cycles. Controlled lab experiments will generate a rich dataset of pressure patterns under varying operating conditions, providing the raw material for training machine learning models.
Once the data is obtained, the scientists will feed the information into machine learning algorithms, training them on these pressure signatures to recognise the subtle cues that indicate the early stages of pre-ignition. The models will be exposed to thousands of combustion cycles, both normal and fault-prone, allowing them to classify engine states with high accuracy.
Once the algorithms perform well in the lab, the next challenge is to adapt them to work with commercially available sensors. These sensors are more affordable and robust but provide lower-resolution data than their lab counterparts. The team will tune the models to effectively interpret the lower-fidelity signals, essential for their translation into the real world.
Download your PDF now!
Implications For Clean Transport
Beyond its potential technical achievement, the project signals a broader opportunity for cleaner, more reliable hydrogen-powered transport. Real-time monitoring using affordable components will open the door to scalable applications in sectors like heavy-duty trucking, where engine durability and safety are non-negotiable.
This work also hopes to move diagnostics from the workshop floor to the road itself, replacing periodic checks and post-failure analysis with continuous, onboard assessment. It brings machine learning into the engine bay and could make smart combustion control a practical reality.
If successful, this approach could form the basis of commercial diagnostic tools for hydrogen engines, contributing to the broader effort to decarbonise transport while maintaining performance and reliability.
Reference
Press Release. SwRI, UT San Antonio collaboration uses machine learning to detect pre-ignition in hydrogen engines. Southwest Research Institute. Accessed on 16 September 2025. https://www.swri.org/newsroom/press-releases/swri-ut-san-antonio-collaboration-uses-machine-learning-detect-pre-ignition-hydrogen-engines