Riassunto analitico
Knock is one of the most damaging and anomalous forms of combustion in spark-ignition internal combustion engines, with the potential to cause significant damage to components such as the piston head, as well as limiting overall engine performance. Various methods for determining the feedback signal for combustion knock have been implemented, with the most common approach being the monitoring of engine vibrations via accelerometers. While traditional techniques perform adequately for commercial engines, they are not always suitable for high-performance engines, where much higher rotational speeds are involved.
At higher RPMs, these engines generate significant vibration, making it difficult to distinguish knock events from the noise components in the accelerometer output. In such environments, knock detection becomes crucial, not only for performance optimization but also for ensuring the longevity of engine components.
This thesis presents the development of an advanced knock recognition system using accelerometers to detect and analyze knock events in high-performance engines. Advanced signal processing techniques were employed to identify knock events with high accuracy across various operational conditions. Starting with traditional knock recognition methods, a novel Knock Index, termed the 'Shock Response Spectrum,' was developed. This index enables better characterization of knock events even under complex vibration conditions.
Additionally, an optimization algorithm was introduced to dynamically select the best angular window and filter frequency for signal processing. This approach ensures maximum differentiation between knock events and noise, improving detection reliability. The system’s performance was evaluated through extensive testing, demonstrating a marked improvement in knock detection accuracy over existing techniques, particularly in high-speed engine scenarios. The results are then analyzed in conjunction with telemetry data, which provides valuable insights into the specific conditions under which knock typically occurs. This integration enables further refinement of the detection process by filtering out false knock detections, which can otherwise skew the analysis and compromise the accuracy of the system. By correlating knock events with engine operating parameters such as load and rotational speed, the system becomes more effective at distinguishing between genuine knock events and noise, thereby enhancing the reliability of the overall analysis.
Furthermore, machine learning algorithms were incorporated to offer a complementary strategy for knock recognition. By training models on engine vibration and telemetry data, these algorithms provide a powerful tool to identify knock patterns under varying operating conditions, potentially adapting to new engines and configurations without extensive manual calibration.
The results of this research show that the proposed system effectively detects knock, even under challenging high-performance engine conditions, and offers significant improvements over traditional detection methods. This work contributes to the development of more robust, accurate, and adaptive knock detection systems, with clear applications in modern high-performance engine control systems, where precision and reliability are paramount.
|
