Development of a real-time integrated measurement system for polar ice load and navigation

2.1 Hull-mounted Hardware Architecture

The onboard platform is configured to acquire in real time and on a common time base—four data domains that govern ship–ice interaction:

• ✔ Local structural response (hull strains)
• ✔ Global hull response (6-DOF motions)
• ✔ Ice-field environment (thickness, visual record)
• ✔ Operating parameters (power, speed, trim, etc.)

2.1.1 Local Ice-load Sensing

Forty-two three-axis resistive strain-gauge rosettes were bonded to the inner surface of the bow shell plating between frames 105 and 110. The term 'inner hull plating' is avoided here because it can be misunderstood as a reference to a separate inner-hull or double-bottom structure. External mounting on the wetted shell surface was not adopted because the gauges and lead wires would then be exposed directly to ice abrasion, coating damage, seawater ingress, and severe maintenance constraints during repeated polar operations. The installed rosettes therefore measure the structural response transmitted through the shell plate from the protected inner surface.

In the uploaded export, the rosettes appear as 126 raw A/B/C channels together with 42 SH and 42 AG derived outputs, all sampled at 50 Hz. Here, SH denotes the in-plane engineering shear strain derived from each 45-degree rosette, SH = 2[ε_B - (ε_A + ε_C)/2], and AG denotes the corresponding principal-strain angle, AG = 0.5 atan2(SH, ε_A - ε_C), reported in degrees and wrapped to 0-180 degrees. These channels measure structural response; they are not interpreted in this paper as direct ice-pressure measurements.

Table 1:

System architecture and data interfaces of the integrated onboard platform

2.1.5 Operating Data Interface

Propulsive and navigation variables such as delivered power, shaft rpm, speed over ground, trim, and draft are interfaced from the shipboard recording systems to provide operational context for later analysis.

Figure 2:

Example of the real-time onboard visualisation interface (top: integrated dashboard view for navigation and contextual video; bottom: multi-channel waveform quick-look view for strain-gauge monitoring)

2.2 Data acquisition and Current Export Structure

The platform employs a dedicated acquisition-and-synchronization module that collects strain-gauge, IMU, EM-thickness, camera, VDR/AMS, and GNSS streams on a shared onboard framework. Table 2 summarizes the current sensor channels and export status used in this revision.

Table 2:

Sensor channel configuration and current export status

In the current implementation, waveform channels are sampled at 50 Hz and stored internally in 30 s NetCDF chunks together with metadata; external exports are generated as CSV for downstream analysis. This distinction between sampling interval and file-chunk duration is retained here because it affects later post-processing and auditability.

2.3 Onboard Data Processing and Visualization

When synchronized sensor streams arrive at the shipboard server they are processed in near-real time through signal decoding, unit conversion and filtering, generation of synchronized response indicators, and screen rendering. The current acquisition and display workflow is implemented through catman AP and associated export tools [12].

Three-axis strain readings from each rosette are stored together with SH and AG derived outputs. A separate 21 x 21 influence-matrix workbook is available for one hull side and can support later matrix-based load reconstruction; a recent design-load-patch-based derivation of such a matrix for Araon was reported separately [8]. To reduce naming ambiguity at the interface between the onboard export and the analysis model, Table 3 lists the one-to-one correspondence adopted in the present study between the sequential DAQ channel labels and the sensor IDs defined in the influence-matrix workbook for one hull side.

Table 3:

Cross-reference between DAQ channel labels and influence-matrix sensor IDs

The two naming systems are not numerically identical, so Table 3 is used as a cross-reference for later matrix-based interpretation. The present paper does not further discuss the origin of the non-contiguous workbook numbering. The starboard side follows the corresponding RK series. This clarification is sufficient for traceable channel-to-sensor correspondence in the present paper, but full pressure back-calculation is still not reported because a fully audited inversion chain, including sign convention and downstream force integration, has not yet been frozen.

2.4 Remote Transfer and Archival Platform

Because raw ice-trial data can reach tens of gigabytes per day, the platform adopts a staged workflow consisting of onboard storage, selective ship-to-shore transfer, and shore-side archival. In the 2024 configuration examined here, higher-bandwidth satellite transfer was not yet treated as a demonstrated capability of the uploaded dataset.

2.5 Environmental Hardening and System Control

The platform is intended for use in cold, humid, and saline polar environments, so the hardware design emphasizes power redundancy, local heating, waterproof encapsulation, and robust cabling.

The measuring network is supplied from the ship AC switchboard through a protected DC conversion stage, and critical instruments can be buffered by UPS (uninterruptible power supply) hardware to reduce data corruption during short disturbances.

Figure 3:

Remote-transfer and shore-side archiving workflow

Outdoor instruments are housed in insulated enclosures with local heating control to mitigate icing and condensation.

External links use cold-resistant Ethernet and sealed penetrations to withstand seawater spray and repeated deck operations.

Further optimization of the thermal-management strategy remains possible, but this topic is outside the scope of the present paper.

3.1 Campaign Context and Representativeness

The representative synchronized segment used for detailed plots in this paper contains 14,857 samples at 50 Hz, corresponding to approximately 297.1 s. To clarify how representative that segment is with respect to the 2024 campaign, the revised assessment also checked continuity across the processed campaign archive. The broader archive contained 41 official ice tests (OT02-OT42), 4,335,303 synchronized context/response rows, and approximately 24.08 h of records after file concatenation. That broader corpus was used only to confirm that synchronized exports existed across the campaign and to identify file-level quality flags; it was not used here to claim campaign-wide performance or load statistics.

3.2 Representative Synchronized Response and Event Level Interpretation

The representative strain-gauge export contains 126 raw A/B/C strain channels, 42 SH channels, and 42 AG channels. The synchronized IMU export contains six numerical channels. No missing numerical samples were found in the representative strain-gauge channels or in the six IMU measurement channels.

Figure 4:

Representative operating area and contextual ice-condition information used during the August 2024 Arctic Voyage

For event screening, the revision uses the sample-wise maximum absolute shear-strain response, S_H,max = max_j |SH_j|, and the accelerometer resultant deviation |a_res - median(a_res)|, where a_res = sqrt(a_x² + a_y² + a_z²). Within the zoomed interval of Figure 6(b) (81.0-85.1 s), S_H,max reached 88.8 microstrain at 83.14 s, whereas the peak accelerometer resultant deviation of 0.047 g occurred at 82.00 s. The two peaks fall within the same short event window but are not coincident. This offset indicates that localized structural bursts and whole-ship motion excursions are related, yet they cannot be treated as a single synchronized scalar peak.

Figure 5 reorganizes the sample-wise SH values at the event time of 83.14 s according to the FE sensor IDs used in the influence matrix. At that instant, the signed SH values ranged from approximately -88.8 to +4.2 µε across the 42 channels. The strongest negative response occurred on the starboard side at FE sensor ID 15 (DAQ channel RK_10), and the next strongest negative response appeared at FE sensor ID 17 (DAQ channel RK_11). Most port-side channels remained within only a few microstrain of zero. This channel-wise pattern is consistent with a localized and spatially non-uniform structural response field. It does not, by itself, determine contact pressure magnitude, exact contact location, or pressure-patch geometry, because those quantities require a fully audited inversion step.

Figure 5:

Event-time SH values arranged by FE sensor ID for subsequent influence-matrix mapping

Figure 6 presents a verified example of synchronized local and global response indicators extracted directly from the uploaded files. Figure 6(a) shows the full-record response, where the left axis represents the maximum absolute SH strain across the available SH channels and the right axis represents the accelerometer resultant deviation derived from the IMU export. Figure 6(b) enlarges the 81.0–85.1 s interval and shows that local strain bursts and the global motion indicator can be compared on the same 50 Hz sample grid within a single synchronized segment. In Figure 6, the accelerometer resultant is a_res = √(a_x² + a_y² + a_z²) (as exported by the IMU, in g-units), and the plotted deviation is |a_res − median(a_res)|. The left axis quantity is max_j |SH_j|, i.e., the maximum absolute shear strain across the 42 SH channels at each sample.

Figure 6:

Verified synchronized local-strain and global-motion indicators extracted from the uploaded segment: (a) full-record time history of the maximum absolute SH strain and the accelerometer resultant deviation; (b) zoomed view of the 81.0–85.1 s interval

The exploratory relation in Figure 7 yields R² = 0.21 between the non-overlapping 2 s mean absolute raw strain and the 2 s peak accelerometer resultant. This weak relation is reported explicitly because it argues against reducing localized hull response and whole-ship motion to a single scalar conversion in the present dataset. The low coefficient of determination indicates that these two screening metrics do not, by themselves, capture the underlying physical relation strongly enough for predictive use. Accordingly, no regression-based load surrogate is proposed here. The figure should therefore be read only as a screening plot showing that the two indicators may rise within the same short event window, while their amplitudes and peak timings are not governed by a simple one-to-one relation. Physically, this is reasonable: the local strain metric depends on contact position and the local structural load path, whereas the accelerometer resultant reflects the rigid-body motion response of the whole ship.

Figure 7:

Exploratory relationship between the 2-s-window mean absolute raw strain and the 2-s-window peak accelerometer resultant in the uploaded segment (R² = 0.21)

Table 4 therefore summarises the representative uploaded files and the current processing status, rather than presenting full-voyage structural-design comparisons.

Table 4:

Representative Segment and Wider Campaign Context

3.3 Practical Implications for Routine Operation

The revised assessment confirms that the current platform can stably export high-dimensional strain-gauge and IMU data on a common sample lattice. At the same time, three implementation-level items remain necessary before routine real-time load inversion should be reported as an operational product: (i) explicit UTC timestamp retention in every export, (ii) finalized DAQ-to-FE sensor mapping with audited sign and side conventions, and (iii) fixed event-tagging rules for selecting comparable impact events.

These remaining tasks do not undermine the value of the present platform. They define the boundary of what can be claimed from the current files. The platform can already be regarded as a practical integrated measurement and data-management system for full-scale monitoring, event screening, and later ship-ice interaction analysis. It should not yet be described as a finalized automatic pressure-inversion product.

Author Contributions

Author Contributions: Conceptualization, J. -S. Ha; Methodology, J. -S. Ha; Software, E. -J. Oh; Validation, Jung-Seok Ha; Formal Analysis, J. -S. Ha; Investigation, J. -S. Ha; Resources, J. -S. Ha; Data Curation, J. -S. Ha; Writing—Original Draft Preparation, J. -S. Ha; Writing—Review & Editing, J. -S. Ha; Visualization, J. -S. Ha; Supervision, J. -S. Ha; Project Administration, S. -Y. Jeong; Funding Acquisition, S. -Y. Jeong.