Generalized Parameter-Adjusted Stochastic Resonance of Duffing Oscillator and Its Application to Weak-Signal Detection

Zhihui Lai, Yong-Gang LengSensors 2015 · OpenReview

TL;DR

This paper introduces a Generalized Parameter-Adjusted Stochastic Resonance (GPASR) model for a Duffing oscillator to overcome the limitations of traditional stochastic resonance (SR) in detecting weak signals. By developing a judgmental function based on the Kramers rate, the authors provide a systematic approach to adjust system parameters when signal characteristics (amplitude, frequency, noise) are mismatched, enabling more robust weak-signal detection in practical engineering applications.

What Problem Does This Paper Address?

Stochastic Resonance (SR) is a phenomenon where a weak signal can be amplified by adding optimal noise to a nonlinear system, making it useful for weak-signal detection. However, a key challenge in applying SR in real-world engineering is the "optimal match" requirement: the signal, noise, and system parameters must align perfectly for SR to occur. In practice, these parameters are often unknown or mismatched. Existing research on parameter-adjusted SR in Duffing oscillators has limitations, focusing mainly on simulation, single-parameter adjustments, and lacking practical engineering applications. This paper addresses this gap by proposing a generalized parameter-adjusted SR (GPASR) model for a Duffing oscillator that systematically analyzes parameter interdependencies and provides clear adjustment rules for various mismatch scenarios, leading to a practical weak-signal detection method.

Methods

The core of this paper is the development of a Generalized Parameter-Adjusted Stochastic Resonance (GPASR) model for a two-dimensional Duffing oscillator. The authors leverage the Kramers rate to establish a judgmental function that quantifies the conditions for SR. This function allows for a systematic analysis of how various parameters (system parameters like damping ratio and potential well shape, signal parameters like amplitude and frequency, and generalized parameters like amplitude and scale transformation coefficients) influence SR.

Key methodological contributions include:

  • Generalized Parameter-Adjusted Duffing Equation: The original Duffing equation is modified with amplitude-transformation coefficient ($\epsilon$) and scale-transformation coefficient ($R$) to account for signal scaling and time/frequency adjustments.
  • Judgmental Function based on Kramers Rate: A function $F(k,a,b, A, f_0, D, \epsilon, R)$ is derived from the Kramers rate condition for SR. SR occurs when $F=1$.
  • Systematic Parameter Analysis: The paper analyzes the monotonic behavior of $F$ with respect to each parameter, identifying rules for adjusting parameters to achieve $F=1$ under different mismatch conditions (unmatched signal amplitude, frequency, and noise intensity).
  • Weak-Signal Detection Model: A practical detection method is proposed based on the GPASR model, involving a step-by-step procedure to adjust parameters and identify weak characteristic signals.

Datasets

Not discussed in the paper. The paper uses simulated data generated from the Duffing oscillator model and experimental data from two specific mechanical fault diagnosis scenarios (rotor shaft-bending and rolling bearing outer ring fault).

Key Insights and Ideas

  • SR is achievable even with parameter mismatches: The GPASR model demonstrates that by systematically adjusting specific parameters, SR can be achieved even when the input signal, noise, and system are not perfectly matched.
  • Parameter interdependencies are crucial: The paper highlights that parameters are not independent. Adjusting one parameter can affect the optimal conditions for others, and the GPASR framework helps navigate these relationships.
  • Specific parameters are key for specific mismatches:
    • Unmatched signal amplitude is best addressed by adjusting the amplitude-transformation coefficient ($\epsilon$).
  • Unmatched signal frequency (especially large frequencies) is best addressed by adjusting the scale-transformation coefficient ($R$).
  • Unmatched noise intensity requires a more comprehensive adjustment of system parameters ($k, a, b$) and generalized parameters ($\epsilon, R$), with specific rules depending on whether the system is under- or over-resonant.
  • Kramers rate provides a powerful analytical tool: Using the Kramers rate to derive a judgmental function offers a theoretical basis for understanding and predicting SR behavior under various parameter conditions.
  • Practical applicability in fault diagnosis: The paper successfully demonstrates the GPASR method's effectiveness in identifying weak fault signals in mechanical equipment, showcasing its real-world utility.

Key Results

  • The paper provides a comprehensive set of parameter adjustment rules for achieving SR in a Duffing oscillator under various mismatch scenarios (signal amplitude, frequency, and noise intensity).
  • Two practical examples of mechanical fault diagnosis (rotor shaft-bending and rolling bearing outer ring fault) demonstrate the effectiveness of the GPASR-based weak-signal detection method. In both cases, the method successfully identified characteristic fault frequencies that were previously obscured by noise or weak signal strength.
  • For the shaft-bending fault, the method identified the fundamental frequency at 28.08 Hz and, with further parameter adjustment, revealed harmonic components indicative of a shaft-bending fault.
  • For the rolling bearing outer ring fault, the method identified a characteristic frequency at 83.1 Hz and its harmonics, confirming the presence of an outer ring fault.

Limitations and Future Work

  • Experimenter Experience: The paper notes that the successful application of the GPASR method relies on the experimenter's experience in recognizing the output state of the system. This makes it less convenient than fully automated adaptive optimization algorithms.
  • Complexity for Inexperienced Users: For inexperienced experimenters, obtaining the SR output might be challenging even if a weak characteristic signal is present.
  • Limited Range for $\epsilon$ and $R$: The paper mentions that $\epsilon$ and $R$ can only be adjusted within a certain range to avoid calculation errors or overflow, implying practical constraints.
  • Future Work: The paper suggests that further research could explore more advanced adaptive optimization algorithms for parameter adjustment and potentially extend the GPASR framework to other nonlinear systems for weak-signal detection. The authors also imply that a more user-friendly interface or automated parameter selection could enhance the method's accessibility.

Summary updated

Abstract

A two-dimensional Duffing oscillator which can produce stochastic resonance (SR) is studied in this paper. We introduce its SR mechanism and present a generalized parameter-adjusted SR (GPASR) model of this oscillator for the necessity of parameter adjustments. The Kramers rate is chosen as the theoretical basis to establish a judgmental function for judging the occurrence of SR in this model; and to analyze and summarize the parameter-adjusted rules under unmatched signal amplitude, frequency, and/or noise-intensity. Furthermore, we propose the weak-signal detection approach based on this GPASR model. Finally, we employ two practical examples to demonstrate the feasibility of the proposed approach in practical engineering application.
BibTeX
@inproceedings{lai2015generalized,
  author = {Zhihui Lai and Yong-Gang Leng},
  title = {{Generalized Parameter-Adjusted Stochastic Resonance of Duffing Oscillator and Its Application to Weak-Signal Detection}},
  booktitle = {Sensors},
  year = {2015},
  url = {https://openreview.net/forum?id=h1W5q6d8IL},
}
Figure