Using deep neural networks to diagnose engine pre-ignition

Kuzhagaliyeva, Nursulu; Thabet, Ali; Singh, Eshan; Ghanem, Bernard

doi:10.1016/j.proci.2020.10.001

Cited by 13 publications

(3 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They investigated data using the synergistic method combining the computing powers of enhanced decision trees and artificial neural networks. Moreover, Kuzhagaliyeva et al [ 42 ] focused their work on the effectiveness of engines. For modern engines, there is a trend of downsizing and boosting it, but it is important at least keep it running efficiently or also increase it.…”

Section: Introductionmentioning

confidence: 99%

A Method of Predicting Wear and Damage of Pantograph Sliding Strips Based on Artificial Neural Networks

Kuźnar

Lorenc

2021

Materials

View full text Add to dashboard Cite

The impact of the pantograph of a rail vehicle on the overhead contact line depends on many factors. Among other things, the type of pantograph, i.e., the material of the sliding strip, influences the wear and possible damage to the sliding strip. The possibility of predicting pantograph failures may make it possible to reduce the number of these kinds of failures. This article presents a method for predicting the technical state of the pantograph by using artificial neural networks. The presented method enables the prediction of the wear and damage of the pantograph, with particular emphasis on carbon sliding strips. The paper compares 12 predictive models based on regression algorithms, where different training algorithms and activation functions were used. Two different types of training data were also used. Such a distinction made it possible to determine the optimal structure of the input and output data teaching the neural network, as well as the determination of the best structure and parameters of the model enabling the prediction of the technical condition of the current collector.

show abstract

Section: Introductionmentioning

confidence: 99%

A Method of Predicting Wear and Damage of Pantograph Sliding Strips Based on Artificial Neural Networks

Kuźnar

Lorenc

2021

Materials

View full text Add to dashboard Cite

show abstract

“…Based on engine vibration during operation, artificial intelligence assessed its degree of wear. In order to detect early ignitions and knocking as a result, the authors of [10] developed an algorithm implementing artificial intelligence. The task of the neural network is to diagnose the combustion process in the engine during operation on the basis of data collected from the controller.…”

Section: Introductionmentioning

confidence: 99%

The use of neural network algorithms for modeling injection doses of modern fuel injectors

Abramek¹,

Osipowicz²,

Mozga³

2021

Combustion Engines

View full text Add to dashboard Cite

The article presents the possibilities of using artificial intelligence methods to model the injection doses of a modern Common Rail (CR) fuel injector. The presented neural network solution belongs to the experimental models known as black boxes in mechatronics. The backpropagation algorithm and its Levenberg-Marquardt expansion were used for the simulation. The analysis showed that there is a good match between the measurements and the computational model. The proposed solution can be used in the processes of diagnosing not only elements of the injection equipment, but also the internal combustion engine. The paper presents the construction and operation of fuel injectors and the important role of precision pairs work.

show abstract

“…For pre-ignition detection, pressure sensors have been essential for research but are usually considered too expensive for widespread mass production, commercial use (for a possible exception, see the Mazda's SkyActiv-X engine [4]). As one alternative to a pressure sensors, datadriven algorithms have also been used to diagnose pre-ignition recently (2020) [5]. As an affordable surrogate for pressure sensor, many researchers have explored using ion sensors [6,7].…”

mentioning

confidence: 99%

Pre-ignition Detection Followed by Immediate Damage Mitigation in a Spark-Ignited Engine

Singh

Dibble

2021

SAE Technical Paper Series

Self Cite

View full text Add to dashboard Cite

Pre-ignition remains a significant bottleneck to further downsizing and downspeeding technologies employed for reducing CO2 emissions in modern turbocharged spark-ignited engines. Pre-ignition, which occurs rarely, may lead to high peak pressures that auto-ignite the entire charge before TDC. The resulting high-pressure oscillations are known as super-knock, leading to sudden and permanent hardware damage to the engine. Over the years, numerous researchers have investigated the stochastic phenomenon's source and concluded that there is a role of lubricant additives, deposits, gasoline properties, and hot surfaces in triggering pre-ignition. No single source has been identified; the research continues. Here, we take a different approach; rather than continue the search for the source(s) of super-knock, we explore mitigating super-knock by detecting pre-ignition early enough to take immediate evasive action. Such evasive action is expected to suppress knock intensity, thereby saving the engine from any permanent damage.In this regard, the current work offers ways to detect pre-ignition (using ion sensors) and then mitigate engine damage by using immediate fuel enrichment. We present three related explorations.In exploration #1, we explore if the occurrence of ions products from the exhaust can warn that the next cycle has a high probability of preignition. For this next cycle, the intake fuel injection can be suspended or increased to operate engine fuel-rich. We find strong ion activity on every cycle. However, there is a weak correlation between the ion signal and pre-ignition occurrence.In exploration #2, an in-cylinder ion-current sensor is used to discover pre-ignition event unfolding during the compression stroke. When such a rare event is detected, more fuel is immediately injected, making the end gas far less reactive and avoiding autoignition and knock. These explorations #1 and #2 were conducted with a DC-based ion sensor. These explorations showed exciting and promising findings. However, our DC-based ion sensors are prone to low signal-to-noise ratio SNR, leading to false positives (unacceptably high number of false positives.)In Exploration #3, the signal-to-noise ratio improvement is explored by replacing the DC-based system with a novel AC-based system. We find the bandpass filtering of the ion signal is key to improved SNR.

show abstract

Using deep neural networks to diagnose engine pre-ignition

Cited by 13 publications

References 14 publications

A Method of Predicting Wear and Damage of Pantograph Sliding Strips Based on Artificial Neural Networks

A Method of Predicting Wear and Damage of Pantograph Sliding Strips Based on Artificial Neural Networks

The use of neural network algorithms for modeling injection doses of modern fuel injectors

Pre-ignition Detection Followed by Immediate Damage Mitigation in a Spark-Ignited Engine

Contact Info

Product

Resources

About