Design and Field Validation of an Offline Synchronized Multi-Sensor DAQ System for Bridge Structural Health Monitoring

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents the design and validation of a high-density SHM data acquisition system tailored for large-scale bridges. The work is grounded in a clear engineering background, features solid hardware implementation, and is supported by comprehensive field data. Nevertheless, the following issues require attention and revision:

1、In Figure 4, the typographical errors "McroSD" and "Timming" should be corrected to "MicroSD" and "Timing", respectively.

2、The paper merely states that "fundamental bending and torsional mode shapes were identified via FDD", yet fails to present the corresponding specific modal parameters, including natural frequencies, damping ratios, and visualizations of the modal shapes.

3、In the caption of Figure 3, the misspelled term "prototipe" should be amended to "prototype".

4、The operational modal analysis (OMA) results of the self-developed system should be compared with those of the Dewesoft commercial system utilized in the preliminary campaign. Specifically, a quantitative comparison of modal parameters between the two systems is required to be supplemented.

Comments for author File: Comments.pdf

Author Response

We would like to express our sincere gratitude for the time and effort you have dedicated to evaluating our work. Your comments and constructive feedback have been useful in improving the quality, clarity, and technical rigor of our manuscript. Please find below our detailed responses to each of your comments. All modifications and additions in the revised manuscript have been highlighted in yellow.

Comments 1: In Figure 4, the typographical errors "McroSD" and "Timming" should be corrected to "MicroSD" and "Timing", respectively.

Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have corrected the typographical errors "McroSD" to "MicroSD" and "Timming" to "Timing" in the updated Figure 4. This change can be found in Section 2.2. Hardware Architecture of the DAQ Nodes, Figure 4.

Comments 2: The paper merely states that "fundamental bending and torsional mode shapes were identified via FDD", yet fails to present the corresponding specific modal parameters, including natural frequencies, damping ratios, and visualizations of the modal shapes.

Response 2: Thank you for pointing this out. We agree with this comment regarding the critical need to present specific modal parameters and visual validations. Therefore, we have introduced a new section dedicated to the Operational Modal Analysis (OMA) results. We have included a table with the identified experimental natural frequencies (Table 5) and the extracted 3D mode shape visualizations. We have deliberately omitted the calculation of damping ratios due to the high statistical uncertainty these values present under non-stationary environmental excitation, prioritizing more deterministic metrics (frequencies and mode shapes) to validate the hardware. These changes can be found in Section 3.3. Operational Modal Analysis, paragraphs 2, 3, and 4.

Comments 3: In the caption of Figure 3, the misspelled term "prototipe" should be amended to "prototype".

Response 3: Thank you for pointing this out. We agree with this comment. We have corrected the misspelled term "prototipe" to "prototype" in the caption of Figure 3. This change can be found in Section 2.2. Hardware Architecture of the DAQ Nodes, Figure 3 caption.

Comments 4: The operational modal analysis (OMA) results of the self-developed system should be compared with those of the Dewesoft commercial system utilized in the preliminary campaign. Specifically, a quantitative comparison of modal parameters between the two systems is required to be supplemented.

Response 4: Thank you for your insightful comment. We completely agree on the importance of robustly validating our custom hardware. However, we have opted not to include a direct quantitative comparison with the Dewesoft system's OMA results. We respectfully justify this decision by clarifying that the preliminary campaign involving the commercial Dewesoft equipment was strictly exploratory; its sole purpose was to evaluate different measurement technologies and empirically derive the fundamental design requirements for our final hardware, not to perform a comprehensive baseline Operational Modal Analysis of the bridge. We have expanded the manuscript to provide an empirical proof of our system's accuracy: the extraction of 3D mode shapes from a dense grid of 118 independent points. As seen in the newly added Figure 20, the mode shapes are geometrically smooth and completely free from phase distortion. This clean geometric deformation is the ultimate empirical validation that both the offline synchronization algorithm and the noise floor of the proposed hardware are good, rendering a baseline comparison redundant. We have added a paragraph to explicitly state this validation rationale in the revised manuscript. This change can be found in Section 3.3. Operational Modal Analysis, final paragraph.

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript develops an offline-synchronized multi-sensor data acquisition system to address the drawbacks of traditional monitoring systems for bridge structural health monitoring (SHM), and validates the system via field tests on a real bridge. The work includes substantial hardware development and field testing. However, multiple deficiencies remain throughout the manuscript, and it does not meet the acceptance criteria for this journal. Major Revision is recommended. Detailed comments are listed as follows:

(1) The technical scheme adopted in this work has been widely implemented and reported in the fields of civil engineering SHM and low-power sensor data acquisition. No disruptive or substantial improvements are proposed regarding the overall hardware architecture, time synchronization scheme and data storage strategy. The authors are advised to clarify the distinctive advantages of the proposed system, add comparisons with existing literature and similar devices, and explicitly highlight the innovations of this work.

(2) A preliminary comparison of four data acquisition architectures is presented, yet only qualitative descriptions of their pros and cons are provided without quantitative performance analysis. In addition, the background noise calibration of sensors under no-load conditions is missing, which fails to verify whether the system meets the predefined technical specifications. Quantitative comparison tables and sensor calibration test data should be supplemented.

(3) Time synchronization is a core performance indicator for offline data acquisition systems. In this manuscript, phase stability is only qualitatively verified using cross-correlation analysis of reference nodes, while the time synchronization error among multiple nodes is not quantified. Operational Modal Analysis (OMA) for bridges requires extremely high time synchronization accuracy. Existing literature has proven that microsecond-level time errors can severely degrade modal identification results. This study only cites the nominal accuracy (±5 ppm) of the DS3231 real-time clock, but does not test the long-term clock drift and synchronization deviation of multiple nodes under practical operating conditions. The effects of temperature, vibration and battery voltage fluctuation on clock accuracy are also not analyzed. Furthermore, time calibration is performed only once before system deployment, and the accumulation law of synchronization errors during days of continuous monitoring is not evaluated. Actual measurements of synchronization accuracy, error analysis and corresponding correction strategies need to be added.

(4) Only basic indicator calculations are conducted for the collected acceleration and displacement data. Complete modal identification and mode shape analysis are not implemented, and the displacement signals are not decomposed into multiple components. The large volume of field measurement data is not fully exploited. It is suggested to conduct in-depth data analysis, and perform modal analysis and operational condition assessment based on structural characteristics.

(5) As a hardware development study, key details about circuit design, PCB layout and mechanical structure design are insufficient. Only the overall state machine workflow of the firmware is introduced, while core logic including DMA transmission, interrupt priority, SD card block writing timing and PSRAM cache scheduling is not elaborated. The authors should supplement detailed hardware design information, simulation or comparative tests, as well as key firmware logic.

(6) The ADXL355 accelerometer and ADS1220 analog-to-digital converter are selected for this system. However, performance benchmarks between these two core chips and other mainstream sensors of the same type are not provided. The rationality and uniqueness of the component selection for bridge micro-vibration and micro-displacement monitoring scenarios are not demonstrated.

(7) The field tests are only carried out under regular traffic loads without comparative test conditions. Meanwhile, synchronous comparative tests between the self-developed system and commercial mature DAQ devices are absent.

(8) Although multiple low-power design techniques are adopted in this work, power consumption of functional modules under different operating modes is not tested. The service life is not calculated according to battery parameters, so the long-term working capability of the device for field monitoring cannot be validated. Power consumption tests under all working modes and battery life evaluation should be supplemented.

(9) When discussing system limitations and future improvements, the authors only mention the inability to access data remotely. Potential problems such as adaptability to extreme environments, long-term service performance and applicability to various scenarios are not fully discussed. The integration of communication modules is merely a preliminary concept without detailed design combined with practical engineering requirements. A comprehensive analysis of system limitations and feasible optimization solutions are required.

(10) Several figures (including Figure 1, Figure 3, Figure 4, Figure 12 and Figure 16) suffer from ambiguous labels, incomplete axis information and missing legends. In addition, inconsistent use of professional terminology, textual errors and non-standard reference formatting are observed throughout the manuscript.

Author Response

We would like to express our sincere gratitude for the time and effort you have dedicated to evaluating our work. Your comments and constructive feedback have been useful in improving the quality, clarity, and technical rigor of our manuscript. Please find below our detailed responses to each of your comments. All modifications and additions in the revised manuscript have been highlighted in yellow.

Comments 1: The technical scheme adopted in this work has been widely implemented and reported in the fields of civil engineering SHM and low-power sensor data acquisition. No disruptive or substantial improvements are proposed regarding the overall hardware architecture, time synchronization scheme and data storage strategy. The authors are advised to clarify the distinctive advantages of the proposed system, add comparisons with existing literature and similar devices, and explicitly highlight the innovations of this work.

Response 1: Thank you for pointing this out. We agree with this comment. We have revised the Introduction to explicitly state our distinctive advantages compared to recent literature. We highlight that our true innovation lies in completely decoupling high-speed data acquisition from energy-intensive wireless transmission, avoiding latency bottlenecks via a PSRAM buffer while utilizing a purely hardware-based (RTC-PPS) offline synchronization method. This change can be found in Section 1. Introduction, paragraph 3.

Comments 2: A preliminary comparison of four data acquisition architectures is presented, yet only qualitative descriptions of their pros and cons are provided without quantitative performance analysis. In addition, the background noise calibration of sensors under no-load conditions is missing, which fails to verify whether the system meets the predefined technical specifications. Quantitative comparison tables and sensor calibration test data should be supplemented.

Response 2: Thank you for your comment. We would like to clarify that the preliminary campaign was explicitly conducted to evaluate and validate the measurement technology that would be utilized for the final system design, rather than to perform a comprehensive quantitative performance analysis of those preliminary architectures. To address the noise calibration concern, rather than performing no-load tests, we have added Table 3 (accelerometers) and Table 4 (ADCs) to strictly compare the rigorous manufacturer specifications of the selected components against other alternatives. These official component specifications validate the metrics obtained in the field analysis and justify that the hardware meets the requirement needed for this specific bridge. These additions can be found in Section 2.3 (Table 3) and Section 2.4 (Table 4).

Comments 3: Time synchronization is a core performance indicator for offline data acquisition systems. In this manuscript, phase stability is only qualitatively verified using cross-correlation analysis of reference nodes, while the time synchronization error among multiple nodes is not quantified. Operational Modal Analysis (OMA) for bridges requires extremely high time synchronization accuracy. Existing literature has proven that microsecond-level time errors can severely degrade modal identification results. This study only cites the nominal accuracy (±5 ppm) of the DS3231 real-time clock, but does not test the long-term clock drift and synchronization deviation of multiple nodes under practical operating conditions. The effects of temperature, vibration and battery voltage fluctuation on clock accuracy are also not analyzed. Furthermore, time calibration is performed only once before system deployment, and the accumulation law of synchronization errors during days of continuous monitoring is not evaluated. Actual measurements of synchronization accuracy, error analysis and corresponding correction strategies need to be added.

Response 3: Thank you for pointing this out. We completely agree. To quantitatively address the time synchronization error, we have added Figure 16, which presents a direct magnified time-history comparison between several independent nodes located at the same cross-section of the bridge. This change can be found in Section 3.1. System Performance and Data Integrity, final paragraphs.

Comments 4: Only basic indicator calculations are conducted for the collected acceleration and displacement data. Complete modal identification and mode shape analysis are not implemented, and the displacement signals are not decomposed into multiple components. The large volume of field measurement data is not fully exploited.

Response 4: Agree. We have shown the measurement data by introducing a new section dedicated entirely to OMA. We have extracted and presented the specific modal parameters (Table 5) and the visualizations of the 3D mode shapes (Figures 19 and 20) to validate the integrity of the data collected across the 118 DOFs. These changes can be found in Section 3.3. Operational Modal Analysis.

Comments 5: As a hardware development study, key details about circuit design, PCB layout and mechanical structure design are insufficient. Only the overall state machine workflow of the firmware is introduced, while core logic including DMA transmission, interrupt priority, SD card block writing timing and PSRAM cache scheduling is not elaborated.

Response 5: Thank you for this observation. We consider that the current explanations regarding the hardware design and the overall state machine workflow are sufficient for the scope and objectives of this paper, which focuses on the system-level validation for SHM. Nevertheless, to provide better clarity and address your concerns, we have improved the explanation of the firmware operation. We have detailed the PSRAM caching scheduling and the core interrupt logic to better illustrate how the system avoids data loss. These changes can be found in Section 2.5. Firmware Architecture.

Comments 6: The ADXL355 accelerometer and ADS1220 analog-to-digital converter are selected for this system. However, performance benchmarks between these two core chips and other mainstream sensors of the same type are not provided.

Response 6: Agree. We have, accordingly, added detailed quantitative benchmark tables to justify our component selection over other mainstream industrial alternatives. Table 3 compares the ADXL355 against the Murata SCA3300 and ST IIS3DWB. Table 4 compares the ADS1220 against the AD7124-4 and the internal STM32 ADC. These changes can be found in Sections 2.3 and 2.4.

Comments 7: The field tests are only carried out under regular traffic loads without comparative test conditions. Meanwhile, synchronous comparative tests between the self-developed system and commercial mature DAQ devices are absent.

Response 7: Thank you for your comment. We clarify that the preliminary campaign with the commercial Dewesoft system was explicitly designed to establish the baseline requirements, rather than to serve as an ongoing parallel measurement. Given the logistical impossibility of deploying 118 wired Dewesoft nodes across the active bridge, the ultimate validation of our self-developed system is presented through the new OMA results.

Comments 8: Although multiple low-power design techniques are adopted in this work, power consumption of functional modules under different operating modes is not tested. The service life is not calculated according to battery parameters, so the long-term working capability of the device for field monitoring cannot be validated.

Response 8: Agree. We explicitly quantify the current draw under each state. These calculations can be found in Section 3.1. System Performance and Data Integrity.

Comments 9: When discussing system limitations and future improvements, the authors only mention the inability to access data remotely. Potential problems such as adaptability to extreme environments, long-term service performance and applicability to various scenarios are not fully discussed.

Response 9: Thank you for pointing this out. We agree with this comment. We have expanded the Discussion section to address the inherent risks of purely offline systems (e.g., silent SD card failures and RTC thermal drifts in multi-month scenarios). We incorporated a future roadmap proposing a hybrid Edge-to-Cloud architecture utilizing a low power module for a daily diagnostic and periodic time calibration against network servers, mitigating these physical risks without sacrificing autonomy. These changes can be found in Section 4. Discussion.

Comments 10: Several figures (including Figure 1, Figure 3, Figure 4, Figure 12 and Figure 16) suffer from ambiguous labels, incomplete axis information and missing legends. In addition, inconsistent use of professional terminology, textual errors and non-standard reference formatting are observed throughout the manuscript.

Response 10: Thank you for pointing this out. We agree with this comment. We have reviewed the entire manuscript, correcting the textual errors, formatting references according to the journal guidelines, and updating some of the mentioned figures to be clearer.

Reviewer 3 Report

Comments and Suggestions for Authors

This paper presents the design and field validation of an offline-synchronized multi-sensor DAQ system for bridge structural health monitoring. The system includes inertial nodes based on ADXL355 MEMS accelerometers and displacement nodes. Local storage and RTC-based synchronization are used to avoid the drawbacks of wired systems and wireless transmission. A field campaign on the Spyckstraße bridge demonstrates the feasibility of the proposed system for ambient vibration measurement, displacement monitoring, and preliminary operational modal analysis. However, several key technical aspects require further clarification.

1. The manuscript proposes a useful DAQ architecture, but the difference from existing low-power wireless or local-storage SHM systems is not sufficiently described. The authors should provide a direct comparison.
2. The manuscript states that the data are suitable for OMA, but only preliminary frequency-domain results are shown. The identified natural frequencies, damping ratios, and representative mode shapes should be reported, to demonstrate the effectiveness for bridge modal identification.
3. Local storage avoids wireless packet loss, but it also prevents real-time system health checks and immediate detection of sensor failure. The authors should discuss how the system would handle long-term deployments, battery degradation, SD-card failure.
4. The authors should enrich the literature review by adding recent references related to sensor-network signal synchronization, energy-efficient data transmission, and wireless sensor network reliability to help clarify the technical motivation for adopting the proposed offline synchronization and local-storage architecture.

[1] A Synchronous Transmission Method for Array Signals of Sensor Network under Resonance Technology; [2] Reverse Synchronous Transmission of Electrical Signals Based on Parallel Injection and Series Pickup; [3] Integrating power assignment into energy-efficient routing in E2E retransmission systems

5. The manuscript states that 65 independent sensor nodes were deployed, but the results section mentions that 38 nodes were retrieved. The authors should clarify whether 65 refers to measurement points/sensors and 38 refers to physical DAQ nodes.

Author Response

We would like to express our sincere gratitude for the time and effort you have dedicated to evaluating our work. Your comments and constructive feedback have been useful in improving the quality, clarity, and technical rigor of our manuscript. Please find below our detailed responses to each of your comments. All modifications and additions in the revised manuscript have been highlighted in yellow.

Comments 1: The manuscript proposes a useful DAQ architecture, but the difference from existing low-power wireless or local-storage SHM systems is not sufficiently described. The authors should provide a direct comparison.

Response 1: Thank you for pointing this out. We agree with this comment. We have revised the Introduction to explicitly highlight the distinctive advantages of our system compared to existing solutions. Specifically, we clarify that our architecture completely decouples high-speed data acquisition from energy-intensive wireless transmission, relies on a purely hardware-based offline synchronization method, and utilizes an intermediate PSRAM buffer to eliminate the unpredictable write latencies and buffer overflows common in standard local-storage loggers. This change can be found in Section 1. Introduction, paragraph 3.

Comments 2: The manuscript states that the data are suitable for OMA, but only preliminary frequency-domain results are shown. The identified natural frequencies, damping ratios, and representative mode shapes should be reported, to demonstrate the effectiveness for bridge modal identification.

Response 2: Thank you for pointing this out. We agree with the critical need to present specific modal parameters and visual validations. Therefore, we have introduced a new section dedicated to OMA. We have included a table with the identified experimental natural frequencies (Table 5) and the extracted 3D mode shape visualizations (Figures 19 and 20). However, we have deliberately omitted the calculation of damping ratios due to the high statistical uncertainty these values present under non-stationary environmental excitation, prioritizing more deterministic metrics (frequencies and mode shapes) to validate the hardware. These changes can be found in Section 3.3. Operational Modal Analysis.

Comments 3: Local storage avoids wireless packet loss, but it also prevents real-time system health checks and immediate detection of sensor failure. The authors should discuss how the system would handle long-term deployments, battery degradation, SD-card failure.

 

Response 3: Thank you for pointing this out. We completely agree. We have expanded the Discussion section to address the inherent risks of purely offline systems (e.g., silent SD card failures and battery degradation). To mitigate these risks in long-term deployments, we incorporated a future roadmap proposing a hybrid Edge-to-Cloud architecture that utilizes a low-power communication module for a daily diagnostic and periodic time calibration against network servers, mitigating these physical risks without sacrificing autonomy. These changes can be found in Section 4. Discussion.

Comments 4: The authors should enrich the literature review by adding recent references related to sensor-network signal synchronization, energy-efficient data transmission, and wireless sensor network reliability to help clarify the technical motivation for adopting the proposed offline synchronization and local-storage architecture.

Response 4: We sincerely thank the reviewer for providing these relevant references. We have integrated all three suggested papers into the Introduction to better frame the severe energy overheads and reliability challenges associated with maintaining continuous phase-locked synchronization and data transmission in WSNs. This strengthens the technical motivation for adopting our offline synchronization and PSRAM-buffered local-storage architecture. This change can be found in Section 1. Introduction, paragraph 3, and the References section ([9], [10], [11]).

Comments 5: The manuscript states that 65 independent sensor nodes were deployed, but the results section mentions that 38 nodes were retrieved. The authors should clarify whether 65 refers to measurement points/sensors and 38 refers to physical DAQ nodes.

Response 5: Thank you for highlighting this discrepancy. We apologize for the previous confusing terminology. We have thoroughly revised the manuscript (Abstract, Section 2.6, and Section 3) to explicitly separate the number of physical hardware nodes manufactured from the number of geometric measurement points covered. To clarify the final scale of the campaign: the deployment utilized a hardware pool of exactly 53 physical sensor nodes (41 inertial + 12 kinematic). By employing a hybrid setup with 26 fixed reference nodes and systematically moving the remaining 15 roving nodes, the system successfully measured a total of 118 distinct geometric points (106 inertial locations and 12 bearing displacements). This clarification can be found in Section 2.6. Field Validation Setup.

Reviewer 4 Report

Comments and Suggestions for Authors

The paper presents an interesting and practically relevant attempt to design an autonomous offline synchronized multi-sensor DAQ system for bridge SHM. The hardware architecture is useful, especially the local storage strategy, RTC-based synchronization, and separated inertial/displacement nodes. And the result is very interesting.  However, I would like to raise several comments and questions about the manuscript.

1. Before we go to the technical parts, I would like to suggest that before you submit the manuscript. A comprehensive proof-read is necessary. There are some typos or grammar mistakes which should be avoided.  

2. The synchronization performance is not demonstrated clearly enough. Since offline synchronization is one of the main contributions, the authors should report the actual achieved synchronization accuracy, clock drift, initial/final offsets, and the cross-correlation procedure. The statement that phase relationships “remained perfectly locked” is too qualitative, and the sentence about “achieving a temporal alignment accuracy” seems incomplete.

3. The paper mentions a preliminary campaign with a commercial IEPE system, but the final proposed nodes are not directly benchmarked against this reference in the results. Colocated comparisons in time domain, frequency domain, coherence, noise floor, amplitude error, phase error, and modal frequencies would make the validation much stronger.

4. Some performance claims are overstated relative to the reported evidence. Claims such as zero data loss, stable battery operation, reliable SD-card burst writing, and robust multi-week deployment need more concrete statistics. For example, the authors should report the number of files/samples, missing packets, counter discontinuities, SD write latency, battery voltage evolution, and actual autonomy. The manuscript should also clarify the inconsistency between 65 deployed nodes and 38 retrieved nodes.

5. The paper describes the system as interrupt-driven with DMA, but later states that the MCU actively polls the sensor FIFO. These are different acquisition strategies and have different implications for jitter and power consumption. Please clarify the exact acquisition loop, time-stamping method, sampling clock, FIFO handling, DMA usage, and how the 1 Hz RTC is linked to the 1000 Hz acceleration sampling.

Author Response

We would like to express our sincere gratitude for the time and effort you have dedicated to evaluating our work. Your comments and constructive feedback have been useful in improving the quality, clarity, and technical rigor of our manuscript. Please find below our detailed responses to each of your comments. All modifications and additions in the revised manuscript have been highlighted in yellow.

Comments 1: Before we go to the technical parts, I would like to suggest that before you submit the manuscript. A comprehensive proof-read is necessary. There are some typos or grammar mistakes which should be avoided.

Response 1: Thank you. We agree and have conducted a proofreading of manuscript to correct typographical and grammatical errors, and to standardize technical terminology.

Comments 2: The synchronization performance is not demonstrated clearly enough. Since offline synchronization is one of the main contributions, the authors should report the actual achieved synchronization accuracy, clock drift, initial/final offsets, and the cross-correlation procedure. The statement that phase relationships “remained perfectly locked” is too qualitative, and the sentence about “achieving a temporal alignment accuracy” seems incomplete.

Response 2: Thank you for this critical observation. We agree that robust quantitative validation is essential. We have added Figure 16, showing a time-history of independent nodes aligned during a traffic event. These additions are located in Section 3.1. System Performance and Data Integrity.

Comments 3: The paper mentions a preliminary campaign with a commercial IEPE system, but the final proposed nodes are not directly benchmarked against this reference in the results. Colocated comparisons in time domain, frequency domain, coherence, noise floor, amplitude error, phase error, and modal frequencies would make the validation much stronger.

Response 3: Thank you for this comment. The preliminary Dewesoft campaign was strictly exploratory to derive fundamental design requirements and deploying a wired array for direct comparison on an active bridge is logistically unfeasible. Instead, we expanded the manuscript to provide empirical proof of accuracy. Since synchronization errors or excessive noise cause severe visual phase distortion in OMA, the geometrically clean 3D mode shapes in the newly added Figures 18 and 19 validate the hardware's precision. This rationale is now stated in Section 3.3. Operational Modal Analysis.

 

Comments 4: Some performance claims are overstated relative to the reported evidence. Claims such as zero data loss, stable battery operation, reliable SD-card burst writing, and robust multi-week deployment need more concrete statistics. For example, the authors should report the number of files/samples, missing packets, counter discontinuities, SD write latency, battery voltage evolution, and actual autonomy. The manuscript should also clarify the inconsistency between 65 deployed nodes and 38 retrieved nodes.

Response 4: We agree and have reinforced our performance claims with concrete statistics:

• Autonomy and battery: Section 3.1 now includes a mathematical energy profile. By quantifying current draw, we demonstrate that the 7800 mAh battery provides 67 days of periodic sampling.
• Zero data loss: We clarified in Section 2.5 that a 1-byte hardware counter embedded in the payload confirmed zero discontinuities. Additionally, a strict byte-count verification triggers a hardware reset to prevent missing bytes during SD write latencies.
• Node inconsistency: We apologize for the confusion and revised Section 2.6 and the Abstract. The deployment utilized exactly 53 physical sensor nodes (fixed and roving) to measure a total of 118 distinct geometric points.

Comments 5: The paper describes the system as interrupt-driven with DMA, but later states that the MCU actively polls the sensor FIFO. These are different acquisition strategies and have different implications for jitter and power consumption. Please clarify the exact acquisition loop, time-stamping method, sampling clock, FIFO handling, DMA usage, and how the 1 Hz RTC is linked to the 1000 Hz acceleration sampling.

Response 5: Thank you for highlighting this. Since our system uses a unified firmware for two different sensing nodes, both strategies are dynamically employed depending on the node type. We have rewritten Section 2.5. Firmware Architecture to clarify this hierarchy:

• 1 Hz RTC Link: Provides an absolute timestamp anchor exclusively at file creation, not for driving high-frequency sampling.
• Node 1 (Inertial - FIFO Polling): Uses the ADXL355's internal 1000 Hz clock. To avoid interrupt context-switching overhead, the MCU polls the FIFO, reading 7 triaxial samples at once, appending a hardware counter, and pushing the block to PSRAM.
• Node 2 (Kinematic - Interrupt Driven): Given the lower sampling rates required, this module uses an interrupt-driven approach triggered by the ADS1220's DRDY pin.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have fully addressed comments concerning figure standardization, chip selection benchmarking, power consumption calculation, operational modal analysis, and system limitations and future prospects. However, several key review requirements have not been adequately resolved. The study only qualitatively elaborates the system’s innovations without quantitative comparisons with similar devices; quantitative comparisons of the preliminary data acquisition architectures and measured no-load noise data of integrated sensors are not supplemented. Only qualitative time-history waveforms are provided to verify synchronization accuracy, lacking microsecond-level error quantification, analyses of long-term clock drift as well as the influences of temperature, vibration and battery voltage, and corresponding error correction strategies. Synchronous parallel measurement comparisons between the self-developed system and commercial DAQ equipment are absent, and underlying hardware and firmware details including PCB layout, DMA transmission and SD card writing timing remain insufficient. Additional quantitative test datasets, comparative charts and timing diagrams for low-level design are needed to improve the demonstration of core performance and support the novelty of this work.

Author Response

We would like to express our gratitude for your rigorous evaluation of our manuscript. We appreciate your insistence on quantitative rigor and technical transparency. Your comments are relevant from a metrological and embedded-systems engineering perspective.

However, we must respectfully clarify the scope of this manuscript. This study was explicitly conceived as a field validation and proof-of-concept for a cost-effective, high-density DAQ architecture applied to Civil Engineering, rather than a laboratory metrology or low-level embedded systems characterization study. Due to logistical, budget, and time constraints inherent to full-scale bridge instrumentation, conducting exhaustive laboratory-controlled environmental tests or parallel commercial baseline deployments during the final campaign was not feasible.

Nevertheless, we have addressed your valid concerns by expanding the theoretical justifications, relying on manufacturer specifications, and adding a new Discussion paragraph to acknowledge these constraints in the manuscript. Please find our detailed responses below.

Comments 1: The study only qualitatively elaborates the system’s innovations without quantitative comparisons with similar devices; quantitative comparisons of the preliminary data acquisition architectures and measured no-load noise data of integrated sensors are not supplemented.

Response 1: We appreciate your perspective. Regarding quantitative comparisons, we respectfully point to the previously added Tables 3 and 4 (Sections 2.3 and 2.4), which benchmark the core components against mainstream alternatives in terms of noise density, resolution, and current consumption.

Regarding the lack of measured no-load noise data (zero-vibration laboratory calibration), we acknowledge this as a limitation. Our validation relies on the well-documented factory specifications of the ADXL355 and our in-situ results. The successful extraction of structural peaksand 3D mode shapes empirically proves that the integrated system's practical noise floor under operational conditions is fully adequate for OMA, even without a prior laboratory calibration. We have added a sentence in Section 4 (Discussion) explicitly acknowledging the absence of no-load laboratory calibration as a limitation of the current study.

Comments 2: Only qualitative time-history waveforms are provided to verify synchronization accuracy, lacking microsecond-level error quantification, analyses of long-term clock drift as well as the influences of temperature, vibration and battery voltage, and corresponding error correction strategies.

Response 2: You are entirely correct that temperature and battery voltage fluctuations affect crystal oscillators. While we could not conduct controlled climate-chamber tests to measure these specific microsecond drifts, we rely on the strict specifications of the DS3231 Temperature-Compensated Crystal Oscillator (TCXO). The ±5 ppm tolerance guarantees a maximum theoretical drift of 0.432 seconds per day across the entire industrial temperature range (-40°C to +85°C).

Comments 3: Synchronous parallel measurement comparisons between the self-developed system and commercial DAQ equipment are absent...

Response 3: We completely agree that a colocated synchronous test would be the standard for validation. However, as explained in our previous response, the preliminary Dewesoft campaign was strictly exploratory. During the final, dense deployment (spanning 118 DOFs), it was logistically and economically impossible to concurrently deploy a high-end commercial wired array across the active highway bridge to serve as a parallel baseline. We respectfully argue that in the context of OMA, the complete absence of spatial aliasing and phase distortion in the experimentally extracted 3D mode shapes serves as empirical proof of signal fidelity and synchronization accuracy, rendering a permanent commercial baseline redundant for the scope of this field study.

Comments 4: ...and underlying hardware and firmware details including PCB layout, DMA transmission and SD card writing timing remain insufficient. Additional quantitative test datasets, comparative charts and timing diagrams for low-level design are needed to improve the demonstration of core performance and support the novelty of this work.

Response 4: We understand your request for low-level design transparency. However, providing exhaustive PCB layout schematics, precise DMA transfer timing diagrams, and SD card block-writing chronograms falls outside the primary scope and target audience of this manuscript, which is centered on the macro-architecture and civil engineering application of the system. We believe that the detailed state-machine description (Section 2.5) and the firmware flowchart adequately explain the data buffering strategy at a conceptual level sufficient for reproducing the system's operational logic. We kindly request your understanding in keeping the focus of the paper on the structural health monitoring outcomes and the overall system architecture.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript has been reasonably revised. 

Author Response

We would like to express our gratitude for your time and constructive feedback during both rounds of review. Your earlier suggestions were beneficial and improved the clarity of our methodology.

We are delighted that you find the manuscript to be reasonably and satisfactorily revised. Thank you once again for your thorough evaluation and for supporting the publication of our research.

Reviewer 4 Report

Comments and Suggestions for Authors

The author answer all of my questions and modified it correspond. It is recommended to publish.

Author Response

We would like to express our sincere appreciation evaluation of our manuscript. We are extremely grateful for your positive assessment of our revisions and for your recommendation to publish. Your insights have been invaluable in ensuring this work meets the high standards of the journal. Thank you for your time and your endorsement.