The Piston Slap Force Reconstruction of Diesel Engine Using WOA-VMD and Deconvolution

Shigong Fan, Yixi Cai, Yunxi Shi, Zongzhen Zhang2024 · DOI

TL;DR

This paper introduces a novel method to accurately reconstruct the piston slap force in diesel engines, a critical but difficult-to-measure source of noise and vibration. By combining Whale Optimization Algorithm (WOA) with Variational Mode Decomposition (VMD) for signal separation and employing a regularized deconvolution technique, the authors achieve a robust estimation of this elusive force.

What Problem Does This Paper Address?

Diesel engines are known for their high torque and efficiency, making them popular in heavy-duty applications. However, they also generate significant noise and vibration, with piston slap being a major contributor. Piston slap occurs when the piston, due to lateral forces and clearances, impacts the cylinder wall. While combustion pressure can be measured, directly measuring the slap force is challenging. Existing methods often rely on complex simulations or are prone to inaccuracies. This paper aims to bridge this gap by developing an experimental and computational approach to reconstruct the piston slap force from readily available vibration data.

Methods

The proposed method involves a three-step process:

  1. Piston Slap Response Extraction using WOA-VMD:
    • Variational Mode Decomposition (VMD): This technique is used to decompose the complex vibration signal from the engine block into several intrinsic mode functions (IMFs), each representing a distinct frequency component. VMD is preferred over Empirical Mode Decomposition (EMD) due to its robustness against noise and frequency aliasing.
  2. Whale Optimization Algorithm (WOA): To overcome the sensitivity of VMD to parameter selection (k and α), the WOA is employed. WOA, inspired by the social behavior of humpback whales, optimizes these parameters to achieve a more accurate decomposition of the vibration signal and better separation of the piston slap response from other noise sources like combustion.
  3. Impulse Response Function (IRF) Acquisition:
    • A hammer test is conducted on the cylinder block to measure its response to an impulse. This allows for the determination of the transfer function (or impulse response function) between the slap force (input) and the measured vibration (output). This step establishes the dynamic relationship of the engine block.
  4. Slap Force Reconstruction via Deconvolution:
    • Deconvolution: This is the core of the reconstruction process. Given the measured vibration signal (output) and the determined impulse response function (transfer path), deconvolution is used to estimate the original slap force (input).
  5. Regularization (Tikhonov and GCV): Deconvolution is an ill-posed problem, meaning small errors in the input data can lead to large errors in the output. To address this, the authors employ Tikhonov regularization, a standard technique for stabilizing solutions to ill-posed inverse problems. They further refine this by using the Generalized Cross-Validation (GCV) method to select the optimal regularization parameter, ensuring a balance between data fitting and solution smoothness.

Datasets

The study utilized experimental data collected from an in-line four-stroke, four-cylinder diesel engine. The engine specifications are provided in Table 1. Vibration signals were recorded using accelerometers and pressure sensors at various engine operating conditions (different RPMs, torques, and injection advance angles). The paper does not introduce a new benchmark dataset but rather uses experimental data from a specific engine setup.

Key Insights and Ideas

  • WOA-VMD for Signal Separation: Optimizing VMD parameters with WOA significantly improves the ability to isolate the piston slap vibration from the complex engine noise, particularly from combustion. This is crucial because piston slap and combustion often occur simultaneously.
  • Deconvolution as a Powerful Tool: While deconvolution is a known technique for inverse problems, its application here, combined with regularization, provides a practical way to estimate the slap force from vibration data.
  • Impact of Operating Conditions: The study demonstrates that piston slap force is significantly influenced by engine speed, torque, and injection advance angle. Higher torque and speed, and a greater injection advance angle, generally lead to increased slap force.
  • Slap Occurs Post-Combustion Peak: Piston slap doesn't occur at the peak of combustion pressure but rather at the maximum velocity of pressure rise (VPR), suggesting a dynamic interplay between combustion events and piston motion.
  • Regularization is Key: The success of the deconvolution method hinges on effective regularization to handle the inherent ill-posedness of the problem, with GCV providing a robust way to tune this regularization.

Key Results

  • The proposed WOA-VMD method successfully decomposes vibration signals into IMFs, with specific IMFs identified as primarily representing piston slap.
  • The deconvolution method, combined with Tikhonov regularization and GCV parameter selection, reconstructs the piston slap force with reasonable accuracy.
  • Figures 12-16 show reconstructed slap force waveforms under various engine load conditions, consistently exhibiting two peaks within a specific crank angle range after top dead center, attributed to intense combustion.
  • The study quantifies how increasing torque and speed, and advancing injection timing, lead to higher slap forces.

Limitations and Future Work

  • Linearity Assumption: The method assumes a linear relationship between the slap force and the resulting vibration on the cylinder block. While this simplifies the problem, real-world engine dynamics can be more complex.
  • Generalizability: While the authors suggest the method can be extended to other engine types (gasoline, hybrid), further validation across a wider range of engines and operating conditions would be beneficial.
  • Noise Sensitivity: Although regularization helps, the method's performance might still be affected by high levels of additive noise in the vibration signals.
  • Future Work: The authors suggest extending this method for predictive maintenance, identifying potential issues in engine components before they lead to failure. They also propose incorporating the findings into engine design to optimize structural characteristics and reduce noise and vibration.

Summary updated

BibTeX
@misc{fan2024piston,
  author = {Shigong Fan and Yixi Cai and Yunxi Shi and Zongzhen Zhang},
  title = {{The Piston Slap Force Reconstruction of Diesel Engine Using WOA-VMD and Deconvolution}},
  year = {2024},
  doi = {10.3390/s24123833},
}
Figure