Research on Dynamic Inertial Estimation Technology for Deck Deformation of Large Ships

Bo Ren, Tianjiao Li, Xiang Li2019 · DOI

TL;DR

This paper introduces a novel dynamic inertial estimation method using a combination of a fixed Inertial Navigation System (INS) and a sliding Sub-Inertial Navigation System (SINS) to accurately measure the real-time deformation of large ship decks. By employing a wavelet-Kalman filter approach for noise and outlier reduction, the method achieves high accuracy and a wider measurement range compared to traditional techniques, making it crucial for maintaining the operational integrity of deck-mounted systems.

What Problem Does This Paper Address?

Large ships, particularly aircraft carriers and long-range survey vessels, are equipped with sensitive weapon systems and launching equipment that are highly susceptible to deck deformation. This deformation, caused by factors like wave motion, ship dynamics, and structural stress, can significantly impair the accuracy and safety of these critical systems. Existing methods for estimating deck deformation, such as optical or photogrammetric techniques, often have limitations in terms of coverage or real-time applicability. While inertial measurement matching methods offer good accuracy, they typically provide only point-wise measurements, leaving large areas of the deck unmonitored. This paper addresses the need for a comprehensive, real-time method to dynamically estimate deck deformation across the entire observing trajectory, enabling timely repairs and ensuring operational readiness.

Methods

The core of this research lies in a dynamic estimation method that combines a fixed main Inertial Navigation System (INS) acting as a benchmark with a sliding Sub-Inertial Navigation System (SINS). The INS, installed at the ship's center, provides reference attitude and position data. The SINS, mounted on the deck, moves along the trajectory, measuring deck angle motion. The difference in angular motion between the INS and SINS is used to derive the true angular deformation of the deck.

To handle the noisy data from these inertial sensors, a hybrid filtering approach is employed:

  • Wavelet Filtering: This technique is used to effectively remove outliers (sudden jumps or anomalies) from the acquired data. The authors specifically chose a 3rd-order Daubechies wavelet, selected for its shape similarity to typical outliers, to decompose and reconstruct the signal, isolating and removing these erroneous points.
  • Kalman Filtering: Following wavelet denoising, a standard discrete Kalman filter is applied to reduce the impact of white noise. This recursive algorithm provides a robust estimation of the system's state by optimally combining predictions with measurements.

The combination of these two filters is applied to both the INS and SINS data. The filtered angular motion data, along with the SINS velocity, are then used with differential geometry principles to calculate the curvature and torsion of the deck deformation.

Datasets

The paper utilizes simulated data for evaluation. The simulation is based on the "Kuznetsov" aircraft carrier, with specific parameters for initial latitude, longitude, ship velocity, SINS accuracy, gyro drift, initial attitude, and stochastic process correlation times. The SINS is simulated to move along a 100-meter trajectory over 20 seconds. No real-world datasets were used for training or evaluation.

Key Insights and Ideas

  • Bridging the Gap: The paper successfully extends the concept of point-wise deck deformation measurement to a continuous trajectory-based estimation, significantly increasing the coverage and utility of inertial measurement techniques.
  • Synergistic Filtering: The combination of wavelet and Kalman filtering proves to be more effective than either method alone. Wavelets excel at outlier removal, while Kalman filters are adept at reducing Gaussian noise, leading to a more accurate and robust estimation.
  • Dynamic Deformation Characterization: By calculating curvature and torsion, the method provides a detailed geometric description of the deck's deformation, allowing for precise identification of problematic areas.
  • Practical Relevance: The simulation results demonstrate that the method can accurately identify "potential danger zones" on the deck, which is crucial for proactive maintenance and safety.

Key Results

  • The proposed dynamic inertial estimation method, using the wavelet-Kalman filter, significantly improves the accuracy of angular velocity measurements compared to raw data or data filtered by Kalman filter or wavelet alone.
  • In simulations, the filtered angular velocity errors (RMS) were reduced by an order of magnitude. For example, the RMS error for the 'x' axial velocity decreased from 1.0157 (original data) to 0.4036 (filtered data).
  • The method accurately identified a "potential danger zone" of approximately 6 meters at the beginning of the simulated trajectory, where curvature and torsion exceeded design thresholds.
  • The filtering algorithm effectively suppressed high-frequency noise while preserving the main frequency component of the angular velocity signal.

Limitations and Future Work

  • Sensor Accuracy Requirement: The paper acknowledges that the proposed method requires extremely high accuracy from the inertial sensors.
  • SINS-Deck Contact: It is also noted that the SINS must closely fit the deck during measurement for accurate results.
  • Model Simplification: The authors suggest that using a more accurate model for the deck deformation (beyond the second-order Markov model used) could lead to even more precise simulation results.
  • Future Work: The authors propose that in the future, more complicated deck situations should be measured, implying a need for testing in more dynamic and complex environments.

Summary updated

BibTeX
@misc{ren2019research,
  author = {Bo Ren and Tianjiao Li and Xiang Li},
  title = {{Research on Dynamic Inertial Estimation Technology for Deck Deformation of Large Ships}},
  year = {2019},
  doi = {10.3390/s19194167},
}

Loading related content...

Figure