Method and system for identifying structural characteristics of bridges

Abstract

Disclosed is a method for identifying structural characteristic(s) of a bridge (102, 504, 608). The method comprises: receiving first set (250) of vibration response data associated with first plurality of traversals of first set (230) of vehicles over the bridge, said data being collected by 5 at least one first sensor (106, 210) arranged in each vehicle amongst first set of vehicles and sent to at least one processor (104, 204); receiving second set (260) of vibration response data associated with second plurality of traversals of second set (240) of vehicles over the bridge, said data being collected by at least one second sensor (220) 10 arranged in each vehicle amongst second set of vehicles and sent to the processor(s); and processing, at the processor(s), the first sets of vibration response data and the second set of vibration response data, for identifying the structural characteristic(s).

Description

[0001] METHOD AND SYSTEM FOR. IDENTIFYING STRUCTURAL CHARACTERISTICS OF BRIDGES

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to methods for identifying at least one structural characteristic of a bridge. Moreover, the present disclosure relates to systems for identifying at least one structural characteristic of a bridge.

[0004] BACKGROUND

[0005] Typically, bridge infrastructure forms an important part of transportation infrastructure. The bridge infrastructure is subject to gradual deterioration during its service life due to factors such as aging, environmental conditions, repeated traffic loading, and material degradation. Thus, the bridge infrastructure requires regular assessment of a condition of bridge structures, to ensure structural safety and long-term reliability. Without timely maintenance and effective monitoring, structural deficiencies can escalate, leading to costly repairs or even catastrophic failures.

[0006] Various techniques are used in practice to assess the condition of bridge structures. Such assessment enables detection of possible deterioration that may occur during service life. Such techniques may involve visual inspection procedures, measurement of structural responses, and analysis of data indicative of structural behaviour under operational loading conditions. Visual inspections can be labour-intensive (when conducted by human personnel) and prone to human error, and therefore may be limited in their ability to detect early-stage structural changes. Some structural health monitoring systems for bridges utilize sensors installed directly on bridge structures. Such systems oftenrequire substantial installation, maintenance, and operational resources, which can limit their large-scale deployment across extensive bridge networks. Since such systems are expensive, they are usually implemented only on select bridges (for example, such as long-span or high-priority bridges). This lack of widespread monitoring increases the risk of undetected structural degradation and raises safety risks and maintenance costs. Furthermore, monitoring solutions that rely on specialised measurement configurations or controlled testing conditions may face practical challenges in routine operational environments.

[0007] Therefore, in light of the foregoing discussion, there exists a need to overcome the aforementioned drawbacks.

[0008] SUMMARY

[0009] The aim of the present disclosure is to provide a method and a system to identify at least one structural characteristic of a bridge. The aim of the present disclosure is achieved by a method and a system that collects vibration response data using instrumented vehicles that traverse over the bridge, and processes such data in order to reliably identify the at least one structural characteristic of the bridge, as defined in the appended independent claims to which reference is made to. Advantageous features are set out in the appended dependent claims.

[0010] Throughout the description and claims of this specification, the words "comprise" , "include", "have", and "contain" and variations of these words, for example "comprising" and "comprises" , mean "including but not limited to", and do not exclude other components, items, integers or steps not explicitly disclosed also to be present. Moreover, the singular encompasses the plural unless the context otherwise requires. In particular, where the indefinite article is used, the specification is tobe understood as contemplating plurality as well as singularity, unless the context requires otherwise.

[0011] BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein:

[0013] FIG. 1 is an illustration of an environment in which a system for identifying at least one structural characteristic of a bridge is implemented, in accordance with an embodiment of the present disclosure;

[0014] FIG. 2 is a block diagram of a system for identifying at least one structural characteristic of a bridge, in accordance with an embodiment of the present disclosure;

[0015] FIG. 3 illustrates a flowchart depicting steps of a method for identifying at least one structural characteristic of a bridge, in accordance with an embodiment of the present disclosure;

[0016] FIG. 4 illustrates a graphical representation of vehicle acceleration responses corresponding to different levels of structural change of a bridge, in accordance with an embodiment of the present disclosure; FIG. 5 illustrates a processing framework for processing a given set of vibration response data, in accordance with an embodiment of the present disclosure;

[0017] FIG. 6 illustrates a processing framework for processing a given set of vibration response data, in accordance with another embodiment of the present disclosure;

[0018] FIG. 7 illustrates a processing framework for extracting changesensitive features from a given set of vibration response data, using Mel-Frequency Cepstral Coefficients (MFCCs)-based feature extraction, in accordance with an embodiment of the present disclosure; and FIG. 8 illustrates a processing framework for classifying a bridge health state, in accordance with an embodiment of the present disclosure.DETAILED DESCRIPTION OF EMBODIMENTS

[0019] The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practising the present disclosure are also possible.

[0020] In a first aspect, the present disclosure provides a method for identifying at least one structural characteristic of a bridge, the method comprising:

[0021] receiving a first set of vibration response data associated with a first plurality of traversals of a first set of vehicles over the bridge, wherein the first set of vibration response data is collected by at least one first sensor arranged in each vehicle amongst the first set of vehicles and sent to at least one processor;

[0022] receiving a second set of vibration response data associated with a second plurality of traversals of a second set of vehicles over the bridge, wherein the second set of vibration response data is collected by at least one second sensor arranged in each vehicle amongst the second set of vehicles and sent to the at least one processor; and processing, at the at least one processor, the first set of vibration response data and the second set of vibration response data, for identifying the at least one structural characteristic.

[0023] The present disclosure provides the aforementioned method which enables reliable identification of the at least one structural characteristic of the bridge using the first and second sets of vibration response data obtained from vehicle traversals, without requiring dedicated sensing infrastructure installed on the bridge. By analysing vibration response data collected across multiple traversals and multiple sets of vehicles,bridge-related structural information (i.e., the at least one structural characteristic of the bridge) can be extracted reliably while reducing influence of vehicle-specific dynamics, road roughness, and environmental variability. This improves robustness and repeatability of structural-characteristic identification and enables scalable monitoring of bridge condition using instrumented vehicles operating during normal traffic. The method is implementable using instrumented vehicles (i.e., vehicles having sensors mounted therein) and processor-based analysis, and permits utilization of diverse processing techniques for structural characteristic identification.

[0024] Throughout the present disclosure, the "structural characteristic" of the bridge refers to any measurable or derivable parameter, indicator, behaviour, or attribute that reflects a structural state, a structural configuration, or a structural performance of the bridge. Identifying the at least one structural characteristic provides a technically meaningful basis for evaluating bridge condition and ensuring long-term infrastructure reliability.

[0025] The first set of vehicles and the second set of vehicles include vehicles which are instrumented (i.e., vehicles which have sensor(s) arranged therein). The at least one first sensor may be arranged on a component of each vehicle amongst the first set of vehicles. Likewise, the at least one second sensor may be arranged on a component of each vehicle amongst the second set of vehicles. The component of any vehicle on which a given sensor is arranged could, for example, be a vehicle body, an axle, a suspension component, or a similar structural part of the vehicle.

[0026] The first set of vehicles comprises one or more vehicles. The second set of vehicles also comprises one or more vehicles. In some embodiments, the first set of vehicles is the same as the second set of vehicles. In some other embodiments, the second set of vehicles and the first set ofvehicles have at least one common vehicle. In yet other embodiments, the first set of vehicles is completely different from the second set of vehicles.

[0027] While traversing the bridge, each vehicle acts as both an exciter that shakes the bridge and a sensing device that measures such shaking. The first set of vibration response data is collected over the first plurality of traversals of the first set of vehicles over the bridge. The first set of vibration response data represents vehicle vibrations as the one or more vehicles (of the first set of vehicles) interact with the bridge during the first plurality of traversals. Likewise, the second set of vibration response data is collected over the second plurality of traversals of the second set of vehicles over the bridge. The second set of vibration response data comprises second sensor data representing vehicle vibrations as the one or more vehicles (of the second set of vehicles) interact with the bridge during the second plurality of traversals. The first set of vibration response data and the second set of vibration response data could comprise time-series response data, which can be sent to the at least one processor in real time, or in a batch. A given set of vibration response data includes a plurality of signal components that are indicative, for example, of vehicle dynamics, road roughness, and bridge vibration. The first set of vibration response data and the second set of vibration response data may correspond to different temporal measurement campaigns, different operating conditions, different vehicle configurations, or different observation periods. In some embodiments, using different pluralities of traversals enables extraction of repeating components attributable to bridge dynamics while reducing influence of vehicle-specific or road-induced components.

[0028] The at least one processor refers to hardware, software, firmware, or a combination of these. The at least one processor is configured to execute one or more data processing operations on the first set ofvibration response data and the second set of vibration response data, for identifying the at least one structural characteristic. The one or more data processing operations may include time-domain analysis, frequency-domain analysis, time-frequency analysis, statistical aggregation, model-based parameter estimation, machine learning inference, or combinations thereof. The at least one processor may be implemented as a single computing unit (for example, a processor of a smartphone, tablet computer, desktop computer, or similar), a remote server, or a distributed computing system.

[0029] As the first set of vibration response data and the second set of vibration response data arise from vehicle-bridge interaction (VBI), structural behavior of the bridge influences such datasets, such that indicators related to the at least one structural characteristic can be derived from the first set and the second set. The at least one processor identifies and analyses such indicators via direct estimation or via indirect inference from one or more intermediate derived indicators, for identifying the at least one structural characteristic.

[0030] Optionally, the at least one processor is configured to:

[0031] extract bridge-related structural information comprising at least one of: dynamic characteristics of the bridge, features sensitive to bridge condition, from the first set of vibration response data and the second set of vibration response data; and

[0032] determine at least one of: a structural characteristic of the bridge corresponding to a given plurality of traversals, a deviation between a structural characteristic of the bridge across the first plurality of traversals and the second plurality of traversals, based on the bridge-related structural information.

[0033] In this regard, processing both sets of vibration response data may enable identification of a structural characteristic based on comparison, correlation, aggregation, or differential analysis between the sets. Theidentification of the at least one structural characteristic is performed more reliably than from a single traversal or a single group of traversals, due to repeatability and cross-set analysis of measurement data. Such a manner of identification also has improved robustness against noise, vehicle variability, and environmental influences. Furthermore, such a manner of identification enables detection of subtle structural changes in the bridge.

[0034] Optionally, the at least one structural characteristic comprises at least one of: a structural change, a structural property, a structural feature. Each of these types of structural characteristics express a condition of the bridge in a well-defined form, enabling further evaluation and / or monitoring.

[0035] In this regard, the "structural change" refers to a variation in the bridge-related structural information, between the first and second sets of vibration response data. Examples of the structural change include, but are not limited to, a change in a bridge modal frequency, a change in time-varying bridge frequency behaviour, and a change in a damagesensitive indicator. The damage-sensitive indicator could be a reconstruction error or an anomaly score, an identified damage ratio, or similar.

[0036] The "structural property" refers to a parameter associated with bridge dynamics or structural performance. Examples of the structural property include, but are not limited to, a natural frequency, a damping-related parameter, a stiffness-related parameter, a resonance-related parameter, and a statistical descriptor of a bridge-induced component derived from multiple traversals.

[0037] The "structural feature" refers to a signature, a pattern, or a feature representation that characterises bridge behaviour. Examples of the structural feature include, but are not limited to, a time-frequency signature of bridge response, a spectral peak distribution indicative ofbridge modal content, a coherence peak pattern associated with bridge-induced components, and a feature set that characterises the bridge response.

[0038] Optionally, the step of processing the first set of vibration response data and the second set of vibration response data comprises: performing a coherence analysis on the first set of vibration response data, for determining a first value of at least one bridge modal frequency;

[0039] performing the coherence analysis on the second set of vibration response data, for determining a second value of the at least one bridge modal frequency; and

[0040] determining a difference between the first value of the at least one bridge modal frequency and the second value of the at least one bridge modal frequency, wherein a structural change is identified when the difference is greater than a predefined threshold.

[0041] It will be appreciated that a bridge modal frequency is a natural frequency associated with a specific vibration mode (first mode, second mode, and the like) of the bridge. A bridge modal frequency corresponding to the first mode of vibration may be called 'fundamental modal frequency' or 'bridge fundamental frequency'. Each bridge has multiple natural frequencies that reflect its vibration characteristics, with the fundamental frequency being the lowest natural frequency of vibration. Optionally, the at least one bridge modal frequency comprises at least the bridge modal frequency corresponding to the first mode of vibration.

[0042] Across multiple traversals of a vehicle over the bridge, bridge-induced components exhibit consistent frequency characteristics, whereas vehicle-specific dynamics and road roughness effects may vary. The coherence analysis may be performed across signals within the first set of vibration response data, in order to identify frequency componentsthat are consistently present and correlated across the first plurality of traversals. Frequency components exhibiting high coherence are indicative of intrinsic bridge dynamic characteristics and are thus interpreted as being associated with bridge modal behaviour. In particular, prominent and stable coherent peaks may correspond to and enable isolation of the at least one bridge modal frequency, including the fundamental modal frequency. The technical effect of performing coherence analysis on the first set is that bridge-related modal frequency components can be isolated from other non-bridge-related vibration components, thereby enabling reliable determination of the first value of the at least one bridge modal frequency.

[0043] Similarly, the coherence analysis may be performed across signals within the second set of vibration response data, in order to identify frequency components that are consistently present and correlated across the second plurality of traversals. The technical effect of performing coherence analysis on the second set is that bridge-related modal frequency components can be isolated from other non-bridge-related vibration components, thereby enabling reliable determination of the second value of the at least one bridge modal frequency.

[0044] A technical effect of applying a similar coherence analysis procedure to both sets of vibration response data is that it ensures consistency, such that any detected difference in modal frequency values is attributable primarily to structural behaviour rather than to processing variability. The difference between the first value and the second value may be an absolute difference, a relative difference, or another deviation metric. Since bridge modal frequency is dependent on structural stiffness and mass distribution, the difference between the first value and the second value being greater than the predefined threshold provides an objective and physics-based indicator of the structural change. The predefined threshold may be selected based on expected measurement variability, environmental effects, statistical confidence intervals, or empiricalcalibration data. In one embodiment, the predefined threshold may be determined using a three-sigma (3o) rule, where o represents a standard deviation of the modal frequency under normal conditions, and the structural change may be identified when the modal frequency deviation exceeds both 3o and approximately 5% of the baseline modal frequency.

[0045] Let us consider a first example where each of the first set of vehicles and the second set of vehicles may comprise one bus. Each of the first plurality of traversals and the second plurality of traversals may comprise 10 traversals of the bus over the bridge. Upon performing the coherence analysis on signals within the first set of vibration response data, the first value of a bridge modal frequency may be determined as 4.2 Hertz (Hz). Likewise, upon performing the coherence analysis on signals within the second set of vibration response data, the second value of the bridge modal frequency may be determined as 3.9 Hertz (Hz). The difference between the first value and the second value is 0.3Hz. When the predefined threshold is, for example, 0.2Hz, the structural change may be detected. The structural change could be a local stiffness reduction (for example, development or propagation of a crack in a girder or deck element), a damage to structural connections (for example, loosening or degradation of bolted / riveted joints), bearing or support issues, and the like.

[0046] Optionally, the steps of applying the coherence analysis are performed by applying a coherence-prominent peak identification (coherence-PPI) technique individually to the first set of vibration response data and to the second set of vibration response data. Applying the coherenceprominent peak identification technique to a given set of vibration response data may comprise applying a peak identification procedure to a coherence spectrum determined from signals within the given set of vibration response data, in order to automatically detect one or more prominent coherence peaks corresponding to the at least one bridgemodal frequency. In some embodiments, pairwise coherence calculations performed between signals associated with different traversals may generate corresponding coherence outputs, and such coherence outputs may be aggregated into a coherence set from which the coherence spectrum is obtained and the at least one bridge modal frequency is identified. The coherence-PPI technique may include evaluating at least one of: peak magnitude, peak prominence, peak persistence, across multiple traversals within the given set of vibration response data; and selecting peaks that satisfy at least one predefined criterion associated with such evaluation.

[0047] Applying the coherence-PPI technique individually to the first set of vibration response data and to the second set of vibration response data ensures that values of the at least one bridge modal frequency are determined using consistent peak-selection rules, thereby reducing subjectivity in peak identification and improving repeatability of frequency estimation. The technical effect of employing coherence-PPI is enhanced robustness and automation in bridge modal frequency determination, particularly in the presence of closely spaced spectral components, measurement noise, or signal variability across traversals. Continuing from the first example, applying coherence-PPI to the first set of vibration response data may identify peaks at 2.1Hz, 4.2Hz, 4.4Hz, and 6.8Hz. Evaluation of the peak magnitude, peak prominence, peak persistence, across the 10 traversals within the first set of vibration response data may determine that the most prominent and stable peak is at 4.2Hz (corresponding to the first value of the bridge modal frequency in an earlier example). Likewise, the most prominent and stable peak may be determined at 3.9 Hz, by applying coherence-PPI to the second set of vibration response data.

[0048] Optionally, the method further comprises identifying non-stationary variations of bridge frequencies during the first plurality of traversalsand the second plurality of traversals of the bridge, based on the first set of vibration response data and the second set of vibration response data. In this regard, identifying the non-stationary variations of the bridge frequencies during a traversal comprises determining a timevarying trajectory (i.e., time-varying values) of the at least one bridge modal frequency over at least a portion of the traversal. This captures time varying (i.e., non-stationary) frequency behaviour of one or more bridge-related frequency components within said traversal. In other words, bridge-related frequency content is tracked as a function of time over at least part of the traversal, thereby identifying frequency variations that may occur due to vehicle-bridge interaction effects, transient excitation, and / or changes in operating conditions during traversal. For example, when a heavy vehicle traverses the bridge, the interaction between the heavy vehicle and the bridge may cause variations in the bridge frequencies during said traversal. Furthermore, identifying the non-stationary variations of the bridge frequencies during a given plurality of traversals comprises aggregating, correlating, or statistically characterising the time-varying frequency behaviour across the traversals in a corresponding set of vibration response data (for example by determining a representative frequency-variation trajectory, a variation range, a variation rate, a stability metric, or similar). In this way, traversal-to-traversal variability can be reduced while preserving time-dependent frequency behaviour, thereby enabling more robust comparison of non-stationary frequency behaviour between the first and second pluralities of traversals and reducing a risk of overlooking structural-change indicators that appear in time-varying bridge-frequency behaviour rather than in a single representative frequency value.

[0049] Continuing from the first example, the time-varying trajectory of the bridge modal frequency over one exemplary traversal amongst the first plurality of traversals may be: t = 0 s — > 4.18 Hz, t = 1 s -» 4.22 Hz, t= 2 s 4.26 Hz, and t = 3 s -> 4.23 Hz. Across the 10 traversals amongst the first plurality of traversals, it may be determined that a typical variation range of the bridge modal frequency is 4.18 Hz to 4.26 Hz. Across next 10 traversals amongst the second plurality of traversals, it may be determined that a typical variation range of the bridge modal frequency is 3.86 Hz - 3.93 Hz. This difference in frequency behaviour across these two pluralities of traversals may indicate the structural change.

[0050] Optionally, the non-stationary variations of the bridge frequencies are identified by employing Improved Multisynchrosqueezing Transform (IMSST). In this regard, IMSST is applied as a time-frequency analysis technique to signals within each set of vibration response data (for example, applied per traversal within each plurality of traversals) to generate a sharpened, high-resolution time-frequency representation (TFR) from which time-varying bridge-frequency behaviour can be identified. Notably, trajectory or ridge-following in the TFR. enables the identification of the time-varying bridge-frequency behaviour. IMSST provides a more focused representation in both time and frequency than some conventional time-frequency approaches, which supports clearer identification of non-stationary frequency content in vehiclebridge interaction signals, even in the presence of measurement variability and noise, and reduces sensitivity to smeared or ambiguous time-frequency representations that may hinder reliable tracking of bridge-frequency behaviour.

[0051] Optionally, the step of processing the first set of vibration response data and the second set of vibration response data comprises employing at least one data processing technique that uses the first set of vibration response data as baseline data and the second set of vibration response data as test data, to identify a structural change. In this regard, the baseline data represents a reference structural state of the bridge while the test data represents a later observed structural state of the bridge.The at least one data processing technique is applied to determine whether the test data deviates from the baseline data in a manner indicative of the structural change. The data processing technique may comprise extracting one or more indicators from the baseline data and from the test data and determining a deviation metric between them, applying a model-based analysis of the vibration response signals, learning a baseline representation from the baseline data and evaluating the test data against that baseline representation, or similar processing operations. The deviation may be quantified in the time domain, frequency domain, time-frequency domain, or in a feature domain derived from the first and second sets of vibration response data. The technical effect of using baseline data and test data in this manner is to enable objective and repeatable detection of the structural change by comparison of vibration-response-derived indicators across measurement campaigns, while reducing sensitivity to traversal-to-traversal variability through use of pluralities of traversals in each dataset.

[0052] In some embodiments, the first set of vibration response data is collected when the bridge is in a healthy state, and the second set of vibration response data is collected at a later time for detecting deviations indicative of damage or other structural change.

[0053] Optionally, the at least one data processing technique comprises at least one of:

[0054] a physics-guided time-domain model configured to analyze time-domain signals in the first set of vibration response data and the second set of vibration response data;

[0055] a feature extraction technique to extract change-sensitive features from the first set of vibration response data and the second set of vibration response data, and a classification technique to classify a bridge health state based on the change-sensitive features;a deep auto-encoder (DAE) configured to detect the structural change to the bridge based on a comparison of short-time data between the first set of vibration response data and the second set of vibration response data.

[0056] In embodiments employing the physics-guided time-domain model, the model analyses the time-domain signals within the first set of vibration response data and the second set of vibration response data, to derive indicators associated with bridge condition. The physics-guided timedomain model computes a first change index from the first set of vibration response data and a second change index from the second set of vibration response data. A difference between the first change index and the second change index enables identification of the structural change. A technical effect of employing the physics-guided time-domain model is improved interpretability and physical grounding, such that structural-change identification is based on physically meaningful indices derived from the first and second sets of vibration response data.

[0057] In embodiments employing the feature extraction technique and the classification technique, the change-sensitive features are extracted from the first and second sets of vibration response data and those features are used to classify a bridge health state. As an example, the change-sensitive features may be raw vehicle acceleration signals. In some embodiments, the raw vehicle acceleration signals may be divided into multiple signal vectors, and only valid segments corresponding to portions of traversals during which the vehicle is fully on the bridge may be selected for subsequent classification. In some embodiments, signal vectors corresponding to a reference structural state (obtained from the first set of vibration response data) and signal vectors corresponding to another structural state (obtained from the second set of vibration response data) may be used as baseline data and test data, respectively. The classification of the bridge health state uses thechange-sensitive features of the first and second sets of vibration response data. The classification technique may use the extracted features to determine whether the bridge health state corresponds to a normal condition or a condition indicative of structural change. A technical effect of employing the feature extraction technique and the classification technique is improved sensitivity and robustness to variability, enabling automated classification of bridge condition. This technical effect is also present under differing traversal conditions and vehicle-specific response variations.

[0058] In embodiments employing the deep auto-encoder (DAE), short-time segments of the vibration response data from the first set and the second set are analysed to determine whether the test data deviates from the baseline behaviour. The DAE may learn a representation of baseline vibration behaviour from the first set of vibration response data and subsequently evaluate short-time data from the second set of vibration response data against that representation, thereby detecting structural change based on deviations between the two datasets. Such deviations may be quantified using a reconstruction error metric, for example a mean squared error between an input representation and a reconstructed representation. A technical effect of employing the DAE is real-time, data-driven change detection, even when explicit damage labels are limited. This supports bridge-condition assessment in implementations where labelled damaged-case data are unavailable or difficult to obtain.

[0059] In some embodiments, the short-time data for the DAE is obtained by dividing vibration response signals into short-time segments using framing and windowing operations. As an example, a windowing function such as a Hann window may be applied to the segments, and the resulting segments may be transformed into time-frequency representations, for example using a short-time Fourier transform, before being provided to the DAE. In some embodiments, the DAE maycomprise an encoder for compressing an input representation into a lower-dimensional latent representation and a decoder for reconstructing the input representation therefrom.

[0060] Optionally, the feature extraction technique is Mel-Frequency Cepstral Coefficients (MFCCs)-based feature extraction. The MFCC-based feature extraction yields a compact feature representation that characterises spectral content of vibration response of the bridge in a manner that is suitable for identifying structural change. MFCC-based feature extraction may comprise transforming a time-domain vibration signal into a frequency-domain representation (for example, via a Fourier transform), applying a filterbank (for example, a mel-spaced filterbank) to obtain band-energy values, applying a logarithmic operation to the band energies, and applying a decorrelating transform (for example, a discrete cosine transform) to generate MFCC features. The MFCC features comprise the change-sensitive features. The technical effect of employing MFCC-based feature extraction is that the resulting MFCC features can emphasise change-sensitive characteristics of the vibration response, including frequency-content variations, while reducing dimensionality and improving robustness to measurement variability. This supports subsequent processing such as comparison between baseline and test feature distributions and / or classification of bridge health state using the extracted MFCC features, with improved computational efficiency relative to higher-dimensional signal representations.

[0061] Optionally, the classification technique is Optimized AdaBoost-linear support vector machine. In this regard, the Optimized AdaBoost algorithm combines multiple weak classifiers to improve classification performance, while the linear support vector machine determines a decision boundary that separates different bridge health states in a feature space defined by the change-sensitive features. The Optimized AdaBoost-linear support vector machine may be trained using featuredata corresponding to known bridge conditions. Subsequently, the trained classifier may be applied to the change-sensitive features extracted from the first and second sets of vibration response data, in order to determine whether the bridge health state corresponds to a normal condition or a condition indicative of the structural change. A technical effect of employing the Optimized AdaBoost-linear support vector machine is improved classification accuracy and robustness in bridge health state identification through ensemble-based optimisation and efficient linear decision boundaries in the feature space.

[0062] Optionally, the at least one structural characteristic comprises at least one structural change, the method further comprising:

[0063] calculating an identified damage ratio (IDR) as a ratio of a number of detections of the at least one structural change of the bridge to a total number of traversals; and

[0064] estimating a damage severity, based on the identified damage ratio. In this regard, traversal-level structural change detection outcomes are aggregated over a plurality of traversals to calculate the IDR.. The technical effect of calculating the IDR is that it provides a normalised and statistically robust measure of occurrence of structural-change detections across repeated traversals, thereby reducing sensitivity to isolated false detections and improving reliability of condition assessment.

[0065] Optionally, the damage severity is estimated based on the IDR, by mapping higher IDR values to higher estimated damage severity and lower IDR values to lower estimated damage severity. Such mapping may use one or more of predetermined thresholds, calibrated ranges, or empirical relationships. In some embodiments, the damage severity may be expressed as a severity level within a predefined damage severity scale, said severity level depending on the IDR. For example, the predefined damage severity scale may comprise severity levels suchas no damage, minor damage, moderate damage, and severe damage. As an example, if a structural change is detected in 8 out of 10 traversals, the identified damage ratio (IDR) is 0.8, which may be mapped to the 'severe damage' severity level according to the predefined damage severity scale. In some other embodiments, the damage severity may be expressed as a quantitative severity value within a predefined severity value range, the quantitative severity value being determined based on the IDR.. For example, the damage severity may be 5, within the predefined severity value range of 0-10.

[0066] The technical effect of estimating damage severity based on IDR is to provide a quantitative or a categorical severity indicator derived from repeated traversal measurements, enabling graded assessment of bridge condition rather than a binary change / no-change output.

[0067] Optionally, the method further comprises:

[0068] generating an alert signal associated with identification of the at least one structural characteristic of the bridge; and

[0069] sending the alert signal to a user device.

[0070] In this regard, the alert signal conveys a result of the identification and is communicated to the user device for enabling further action based on the result. The alert signal may be configured to convey information relevant to bridge condition assessment, maintenance planning, or operational decision-making. Optionally, the alert signal comprises at least one of: a notification of the at least one structural characteristic, a maintenance action to be undertaken corresponding to the at least one structural characteristic, maintenance-prioritization information indicating a priority of the maintenance action, a damage progression indication in case the at least one structural characteristic comprises structural damage. A technical effect of generating the alert signal is that it provides an actionable output for a user, thereby enabling timelyawareness and response to the identified at least one structural characteristic.

[0071] Sending the alert signal to the user device enables remote communication of the identified at least one structural characteristic to a human operator, monitoring system, or a software agent. This improves responsiveness and operational integration, such that maintenance personnel, other users, or software agents can receive bridge-condition-related information in a timely manner and undertake corresponding inspection, monitoring, or maintenance measures. The user device may be, for example, a mobile phone, a tablet computer, a laptop computer, a desktop computer, or another device associated with bridge monitoring or maintenance management.

[0072] Optionally, the method further comprises:

[0073] collecting sensor data from at least one third sensor arranged in each vehicle amongst the first set of vehicles and the second set of vehicles; detect when a given vehicle amongst the first set of vehicles and the second set of vehicles traverses over the bridge, by processing the sensor data; and

[0074] activating the at least one first sensor for collecting the first set of vibration response data and activating the at least one second sensor for collecting the second set of vibration response data, only while it is detected that the given vehicle traverses over the bridge.

[0075] In this regard, the at least one third sensor is used to determine when the given vehicle traverses over the bridge, such that collection of corresponding vibration response data by the corresponding sensor(s) is initiated for said traversal. In some embodiments, detection of when the given vehicle traverses over the bridge may be based on detection of the given vehicle entering a predefined bridge region and / or exiting the predefined bridge region. In this regard, the at least one first sensor and the at least one second sensor are activated upon detecting entryof the given vehicle into the predefined bridge region and deactivated upon detecting exit of the given vehicle from the predefined bridge region. In some embodiments, such activation and deactivation may be implemented using one or more control signals generated by the at least one processor. The technical effect of this controlled activation of the at least one first sensor and the at least one second sensor is that bridge-relevant vibration response data can be selectively acquired for ongoing traversals of the given vehicle over the bridge while reducing collection of irrelevant data when the vehicle is not traversing the bridge. This reduces unnecessary data volume, power consumption, and processing burden, while improving the likelihood that the collected first and second sets of vibration response data correspond to vehicle-bridge interaction during traversal.

[0076] Optionally, the at least one third sensor comprises at least one of: a location sensor (for example, a GPS receiver, a GNSS receiver, or similar), an RTK-GPS module, an inertial navigation system, a geofencing module, a proximity sensor which activates on coming in proximity with a corresponding sensor arranged on the bridge. The sensor data collected by the at least one third sensor may indicate that the given vehicle is within a predefined bridge location region, follows a predefined route segment corresponding to the bridge, satisfies a proximity condition associated with the bridge, or meets similar conditions indicating that the given vehicle is traversing the bridge.

[0077] In some embodiments, the sensor data collected by the at least one third sensor may further enable spatial referencing of the first set of vibration response data and the second set of vibration response data along the bridge traversal, such that an identified structural characteristic may be associated with a corresponding bridge region. In this regard, the at least one processor may be configured to associate portions of a given set of vibration response data with corresponding spatial positions of the given vehicle during traversal of the bridge, andto analyse the spatially referenced given set of vibration response data to determine a bridge region (along the traversal) in which indicators of the structural characteristic are observed.

[0078] In a second aspect, the present disclosure provides a system for identifying at least one structural characteristic of a bridge, the system comprising at least one processor communicably coupled to at least one first sensor and at least one second sensor, wherein the at least one processor is configured to:

[0079] receive, from the at least one first sensor, a first set of vibration response data associated with a first plurality of traversals of a first set of vehicles over the bridge, wherein the at least one first sensor is arranged in each vehicle amongst the first set of vehicles;

[0080] receive, from the at least one second sensor, a second set of vibration response data associated with a second plurality of traversals of a second set of vehicles over the bridge, wherein the at least one second sensor is arranged in each vehicle amongst the second set of vehicles; and

[0081] process the first set of vibration response data and the second set of vibration response data, to identify the at least one structural characteristic.

[0082] The present disclosure also relates to the system as described above. Various embodiments and variants disclosed above, with respect to the aforementioned method, apply mutatis mutandis to the system.

[0083] The present disclosure provides the aforementioned system which enables reliable identification of the at least one structural characteristic of the bridge through coordinated acquisition and processing of vibration response data obtained from vehicles traversing the bridge. By communicably coupling the at least one processor with the at least one first sensor and the at least one second sensor arranged in vehicles, the system facilitates distributed sensing using instrumented vehicles whileenabling data processing at the at least one processor. This architecture enables efficient collection, transmission, and processing of vibration response data associated with multiple traversals, thereby improving reliability of identification of the at least one structural characteristic while reducing influence of vehicle-specific dynamics, road roughness, and environmental variability. Furthermore, the system enables scalable deployment for bridge condition monitoring without requiring permanent sensing infrastructure installed on the bridge, while allowing integration of diverse processing techniques at the at least one processor for identifying the at least one structural characteristic.

[0084] Optionally, the system further comprises the at least one first sensor configured to measure the first set of vibration response data, and the at least one second sensor configured to measure the second set of vibration response data. In this embodiment, the at least one first sensor and the at least one second sensor are also a part of the system and are configured to measure vibration response of vehicles traversing the bridge, for generating the first set of vibration response data and the second set of vibration response data, respectively. Such a system provides an integrated sensing-and-processing architecture in which acquisition of vibration response data and identification of the at least one structural characteristic are coordinated within a common technical framework, thereby improving consistency of data acquisition across traversals and reducing interfacing complexity between sensing and processing components.

[0085] Optionally the system further comprises at least one third sensor. A technical effect of including the at least one third sensor in the system is that it enables automated detection of when a vehicle traverses the bridge, thereby allowing controlled activation of the at least one first sensor and the at least one second sensor during bridge traversal. This improves relevance of the collected vibration response data by ensuring that measurements correspond to vehicle-bridge interaction duringtraversal, while reducing collection of irrelevant data when the vehicle is not traversing the bridge.

[0086] Optionally, the system further comprises a memory communicably coupled to the at least one processor, wherein at least the first set of vibration response data and the second set of vibration response data are stored in the memory. The memory may be implemented as at least one of: a local memory associated with the at least one processor (for example, volatile memory such as RAM or non-volatile memory such as flash memory), a storage device associated with a computing unit, a database, or a remote / cloud-based storage system communicably coupled to the at least one processor. A technical effect of providing the memory is that the first set of vibration response data and the second set of vibration response data can be reliably stored and accessed for processing. This facilitates efficient data management and enables repeatable processing of the first and second sets of vibration response data for identifying the at least one structural characteristic of the bridge.

[0087] Optionally, the at least one first sensor is arranged in an axle that couples a pair of wheels of each vehicle amongst the first set of vehicles, and wherein the at least one second sensor is arranged in an axle that couples a pair of wheels of each vehicle amongst the second set of vehicles. The axle is a structural component that directly transmits loads between the wheels and the vehicle body during vehicle motion and during interaction with a surface of the bridge. The axle is located close to wheel-bridge contact interface where excitation forces are transmitted. As a result, vibration signals measured at the axle reflect dynamic responses arising from vehicle-bridge interaction. Such placement of the at least one first sensor and the at least one second sensor enhances a quality and relevance of the corresponding vibration response data, thereby improving reliability of identification of the at least one structural characteristic of the bridge.Optionally, the at least one first sensor and / or the at least one second sensor comprises at least one of: an accelerometer, a motion sensor, a vibration sensor, a displacement sensor. A technical effect of supporting different sensor types is that the system can flexibly employ sensors suited to different vehicle configurations and measurement requirements while still obtaining vibration response data indicative of vehicle-bridge interaction. This improves adaptability of the system across different vehicle platforms and sensing setups while enabling reliable acquisition of vibration response data for identifying the at least one structural characteristic of the bridge.

[0088] The accelerometer may be configured to measure acceleration of a vehicle component, thereby capturing dynamic responses associated with vehicle-bridge interaction. The accelerometer could be a singleaxis accelerometer or a multi-axis accelerometer. The motion sensor may provide motion-related information including acceleration, angular velocity, or orientation changes during traversal. The motion sensor could, for example, be an inertial measurement unit (IMU), a gyroscope, an inclinometer. The vibration sensor may be configured to measure vibration signals generated due to interaction between a given vehicle and the bridge. The vibration sensor could, for example, be a microphone, a contact vibration pickup, a piezoelectric vibration sensor. The displacement sensor may be configured to measure displacement or relative movement of a vehicle component caused by bridge-induced vibrations.

[0089] DETAILED DESCRIPTION OF THE DRAWINGS

[0090] Referring to FIG. 1, illustrated is an environment 100 in which a system for identifying at least one structural characteristic of a bridge 102 is implemented, in accordance with an embodiment of the present disclosure. The system comprises at least one processor (depicted as a processor 104) that is communicably coupled to at least one firstsensor (depicted as first sensor 106) and at least one second sensor (shown in FIG. 2). The first sensor 106 is shown to be arranged in a vehicle 108 amongst a first set of vehicles. The first sensor 106 collects vibration response data 110 associated with at least one traversal of the vehicle 108 over the bridge 102, and sends the vibration response data 110 to the at least one processor 104. The vibration response data 110 is part of a first set of vibration response data associated with a first plurality of traversals of the first set of vehicles over the bridge 102. Similarly, the at least one second sensor is arranged in a second set of vehicles (shown in FIG. 2). The at least one second sensor collects a second set of vibration response data associated with a second plurality of traversals of a second set of vehicles over the bridge 102. The at least one processor 104 processes the first set of vibration response data and the second set of vibration response data, to identify at least one structural characteristic of the bridge 102.

[0091] Referring to FIG. 2, illustrated is a block diagram of a system 200 for identifying at least one structural characteristic of a bridge, in accordance with an embodiment of the present disclosure. The system 200 comprises at least one processor (depicted as a processor 204) communicably coupled to at least one first sensor 210 and at least one second sensor 220. The at least one first sensor 210 is arranged in each vehicle amongst a first set 230 of vehicles. Likewise, the at least one second sensor 220 is arranged in each vehicle amongst a second set 240 of vehicles. The at least one first sensor 210 is configured to measure a first set 250 of vibration response data, and the at least one second sensor 220 is configured to measure a second set 260 of vibration response data. The first set 250 of vibration response data is associated with a first plurality of traversals of the first set 230 of vehicles over the bridge. The second set 260 of vibration response data is associated with a second plurality of traversals of the second set 240 of vehicles over the bridge. The at least one processor 204 processesthe first set 250 of vibration response data and the second set 260 of vibration response data, to identify the at least one structural characteristic of the bridge. As an example, the at least one first sensor 210 is shown to comprise an accelerometer 212. As another example, the at least one second sensor 220 is shown to comprise an accelerometer 222 and a vibration sensor 224.

[0092] Optionally, the system 200 further comprises at least one third sensor 270 arranged in each vehicle amongst the first set 230 of vehicles and at least one third sensor 280 arranged in each vehicle amongst the second set 240 of vehicles. Optionally, in this regard, the at least one processor 204 is configured to: collect sensor data 272 and 282 from the at least one third sensor 270 and 280, respectively; detect when a given vehicle amongst the first set 230 of vehicles and the second set 240 of vehicles traverses over the bridge, by processing the sensor data 272 and 282; and activate the at least one first sensor 210 for collecting the first set 250 of vibration response data and activate the at least one second sensor 220 for collecting the second set 260 of vibration response data, only while it is detected that the given vehicle traverses over the bridge. The at least one processor 204 is configured to generate and utilize control signals 290a and 290b for such activation and consequent deactivation (when the given vehicle is not traversing over the bridge).

[0093] Optionally, the at least one processor 204 generates an alert signal 290 associated with identification of the at least one structural characteristic of the bridge, and sends the alert signal 290 to a user device 292. The at least one processor 204 is communicably coupled to the user device 292.

[0094] Referring to FIG. 3, illustrated is a flowchart depicting steps of a method for identifying at least one structural characteristic of a bridge, in accordance with an embodiment of the present disclosure. At step 302,a first set of vibration response data associated with a first plurality of traversals of a first set of vehicles over the bridge, is received. The first set of vibration response data is collected by at least one first sensor arranged in each vehicle amongst the first set of vehicles and sent to at least one processor. At step 304, a second set of vibration response data associated with a second plurality of traversals of a second set of vehicles over the bridge, is received. The second set of vibration response data is collected by at least one second sensor arranged in each vehicle amongst the second set of vehicles and sent to the at least one processor. At step 306, the first set of vibration response data and the second set of vibration response data are processed, at the at least one processor, to identify the at least one structural characteristic.

[0095] Referring to FIG. 4, illustrated is a graphical representation 400 of vehicle acceleration responses corresponding to different levels of structural change of a bridge, in accordance with an embodiment of the present disclosure. In this example, the structural change may correspond to structural damage of the bridge, and the different levels of structural change may correspond to damage severities. The graphical representation 400 represents acceleration difference (in m / s^) on Y-axis and time on X-axis, for multiple damage severities (10%, 20%, 30%, 40%, and 50%). Each curve represents a vibration response signal measured during traversal of a vehicle over the bridge, and shaded regions (i.e., hatched regions) illustrate variation in the acceleration response signals associated with the respective damage severities. Each shaded region extends between a baseline acceleration level (0E+0) and corresponding high-frequency oscillations of the vibration response signal. The graphical representation 400 shows that vehicle acceleration responses differ with changes in damage severity, for example in terms of signal amplitude and response profile over time. As an example, an amplitude envelope corresponding to 50% damage (represented as light dotted hatch between 0E+0 and correspondinghigh-frequency oscillations) is larger than an amplitude envelope corresponding to 10% damage (represented as a darker dense dotted hatch between 0E+0 and corresponding high-frequency oscillations). Such acceleration response signals may be analysed using a physics-guided time-domain model, feature extraction and classification, or other data-processing techniques. In some embodiments, variation of the acceleration response signals with damage severity may be used to estimate a severity level of the structural change or to determine a damage-sensitive indicator indicative of bridge condition.

[0096] Referring to FIG. 5, illustrated is a processing framework 500 for processing a given set of vibration response data, in accordance with an embodiment of the present disclosure. The given set of vibration response data is collected while a given set of vehicles (depicted for example as a vehicle 502) traverses multiple times over a bridge 504.

[0097] The processing framework 500 shows how coherence analysis is performed across signals from a given plurality of traversals (depicted as Run 1, Run 2, Run 3, ... Run n) within the given set of vibration response data. The coherence analysis is performed to identify frequency components that are consistently present across the given plurality of traversals. In the coherence analysis, pairwise coherence calculations are performed between signals associated with different traversals. Each pairwise coherence calculation produces a corresponding coherence output (depicted for example as Coh 1 {X^}, Coh 2 {X2}, ... Coh n {Xn}, Coh n+1 {Xn+i}, Coh n+2 {Xn+2}, ... and so on) collectively referred to as coherence outputs 506. The coherence outputs 506 are aggregated to form a coherence set 508 represented as [X] = { Xi, X2, ...

[0098]

[0099] > where N = Cn2corresponds to a total number of pairwise coherence calculations among the given plurality of traversals. The coherence analysis determines a coherence spectrum representing the degree of correlation between frequency componentspresent in the vibration response signals obtained during the plurality of traversals. Frequency components that exhibit high coherence across multiple traversals indicate vibration characteristics that are consistently excited during vehicle-bridge interaction and are therefore attributable to intrinsic dynamic behaviour of the bridge rather than to vehicle-specific dynamics or road roughness effects. The coherence analysis may be used for determining a value of at least one bridge modal frequency. In particular, prominent peaks in the coherence spectrum correspond to frequencies that remain stable and correlated across the plurality of traversals. Such prominent peaks therefore indicate the at least one bridge modal frequency. Furthermore, a coherence-prominent peak identification (coherence-PPI) technique may be applied to the given set of vibration response data (based on the coherence set 508), to identify the at least one bridge modal frequency. Referring to FIG. 6, illustrated is a processing framework 600 for processing a given set of vibration response data, in accordance with another embodiment of the present disclosure. The given set of vibration response data may be a first set of vibration response data and / or a second set of vibration response data. In an example, the first set of vibration response data may be used as baseline data and the second set of vibration response data may be used as test data. Vibration response signals 602 in the given set of vibration response data correspond to vehicle-bridge interaction, and are obtained from at least one given sensor 604 arranged in one or more vehicles 606 traversing a bridge 608.

[0100] The vibration response signals 602 are first subjected to framing and windowing operations 610, in which said signals are divided into short-time segments. In some embodiments, a Hann window may be applied to each segment to reduce spectral leakage and improve representation of frequency content. The framed vibration response signals are then processed using a short-time Fourier transform (STFT) 620 to obtain atime-frequency representation of the signals. The resulting timefrequency representations are then provided as inputs to a deep autoencoder (DAE) 630 comprising an encoder 640 and a decoder 650.

[0101] The encoder 640 comprises multiple hidden layers that progressively compress the input representation into a lower-dimensional latent representation, while the decoder 650 reconstructs the input representation from the latent representation.

[0102] The deep auto-encoder 630 may learn a representation of baseline vibration behaviour corresponding to a reference structural state of the bridge 608. Subsequently, vibration response signals corresponding to a later structural state of the bridge 608 may be processed through the trained auto-encoder. A reconstruction error, for example a mean squared error (MSE) 660 may be computed between the input representation and the reconstructed representation. A large reconstruction error indicates that the vibration response signals deviate from the learned baseline behaviour. In this regard, the deep auto-encoder 630 may be configured to detect the structural change to the bridge 608 based on a comparison of short-time data between the first set of vibration response data and the second set of vibration response data, wherein deviations between baseline behaviour and later observed behaviour indicate the structural change.

[0103] Referring to FIG. 7, illustrated is a processing framework 700 for extracting change-sensitive features from a given set of vibration response data, using Mel-Frequency Cepstral Coefficients (MFCCs)-based feature extraction, in accordance with an embodiment of the present disclosure. In this example, signals in the given set of vibration response data correspond to vehicle acceleration signals 710 obtained from at least one given sensor arranged in a given set of vehicles traversing a bridge. The vehicle acceleration signals 710 are collected during a given plurality of traversals of the given set of vehicles over the bridge. The vehicle acceleration signals 710 are first subjected todata preprocessing operations 720, which may include, for example, framing and windowing of the vehicle acceleration signals 710. In an example, the vehicle acceleration signals 710 may be divided into short-time segments, and a windowing function may be applied to each segment to reduce spectral leakage and improve representation of frequency content. Next, the vehicle acceleration signals 710 are processed using a Fast Fourier Transform (FFT) 730 to obtain a frequency spectrum representation of each short-time segment. The frequency spectrum represents the distribution of signal energy across different frequency components. The frequency spectrum is then processed through a Mel filterbank 740, which applies a set of filters spaced according to a mel frequency scale. The Mel filterbank 740 produces band-energy values representing energy levels within multiple mel-scaled frequency bands. Next, the band-energy values are subjected to a logarithmic operation 750, for example applying a function of the form y = ln(x), to obtain log-band energies. The logarithmic transformation improves representation of relative energy variations across the frequency bands. The log-band energies are then processed using a Discrete Cosine Transform (DCT) 760, which performs a decorrelating transformation on the log-band energy values to produce a compact feature representation. The resulting features correspond to Mel-Frequency Cepstral Coefficients (MFCCs) 770. In this regard, the MFCCs 770 comprise the change-sensitive features extracted from the given set of vibration response data, which characterise spectral properties of the given set of vibration response data. The MFCC-based feature extraction process illustrated in FIG. 7 therefore generates a compact feature representation suitable for identifying a structural change of the bridge. In some embodiments, the MFCCs 770 may be used as inputs to a classification technique configured to classify a bridge health state based on the changesensitive features.Referring to FIG. 8, illustrated is a processing framework 800 for classifying a bridge health state 850, in accordance with an embodiment of the present disclosure. In this example, vehicle vertical acceleration signals 810 correspond to signals in the first and second sets of vibration response data. The vehicle vertical acceleration signals 810 may be subjected to signal preprocessing operations 820. In this stage, vehicle vertical acceleration signals 810 corresponding to different structural states of a bridge may be divided into multiple signal vectors 822a and 822b. For example, vibration response signals corresponding to a reference structural state of a bridge may be divided into Nh vectors denoted as 822a, while vibration response signals corresponding to another structural state of the bridge may be divided into Nd vectors denoted as 822b. Here, Nh and Nd are numbers of vectors in 822a and 822b, respectively. In some embodiments, only valid acceleration segments corresponding to portions of traversals during which a vehicle is fully on the bridge are selected, so that the resulting signal vectors 822a and 822b represent vehicle-bridge interaction responses. The resulting signal vectors 822a and 822b constitute change-sensitive features derived from the vehicle vertical acceleration signals 810. These change-sensitive features are then provided as inputs to a classification technique 830 configured to classify the bridge health state 850 based on the change-sensitive features. In some embodiments, the classification technique 830 is an Optimized AdaBoost-linear support vector machine 840, which combines ensemble boosting with a support vector machine classifier to determine whether the bridge health state 850 corresponds to a normal condition or a condition indicative of structural change. In this regard, the signal vectors 822a may be used as baseline data, while the signal vectors 822b may be used as test data. The classification technique 830 processes the change-sensitive features extracted from thebaseline data and the test data to determine whether the structural change of the bridge has occurred.

[0104] Provided figures are merely examples, which should not unduly limit the scope of the claims herein. A person skilled in the art will recognize many variations, alternatives, and modifications of embodiments of the present disclosure.

Claims

CLAIMS1. A method for identifying at least one structural characteristic of a bridge (102, 504, 608), the method comprising:receiving a first set (250) of vibration response data associated with a first plurality of traversals of a first set (230) of vehicles over the bridge, wherein the first set of vibration response data is collected by at least one first sensor (106, 210) arranged in each vehicle (108) amongst the first set of vehicles and sent to at least one processor (104, 204);receiving a second set (260) of vibration response data associated with a second plurality of traversals of a second set (240) of vehicles over the bridge, wherein the second set of vibration response data is collected by at least one second sensor (220) arranged in each vehicle amongst the second set of vehicles and sent to the at least one processor; andprocessing, at the at least one processor, the first set of vibration response data and the second set of vibration response data, for identifying the at least one structural characteristic.

2. A method according to claim 1, wherein the at least one structural characteristic comprises at least one of: a structural change, a structural property, a structural feature.

3. A method according to any of the preceding claims, wherein the step of processing the first set (250) of vibration response data and the second set (260) of vibration response data comprises:performing a coherence analysis on the first set of vibration response data, for determining a first value of at least one bridge modal frequency;performing the coherence analysis on the second set of vibration response data, for determining a second value of the at least one bridge modal frequency; anddetermining a difference between the first value of the at least one bridge modal frequency and the second value of the at least one bridge modal frequency, wherein a structural change is identified when the difference is greater than a predefined threshold.

4. A method according to claim 3, wherein the steps of applying the coherence analysis are performed by applying a coherence-prominent peak identification (coherence-PPI) technique individually to the first set (250) of vibration response data and to the second set (260) of vibration response data.

5. A method according to any of claims 3 or 4, further comprising identifying non-stationary variations of bridge frequencies during the first plurality of traversals and the second plurality of traversals of the bridge, based on the first set (250) of vibration response data and the second set (260) of vibration response data.

6. A method according to claim 5, wherein the non-stationary variations of the bridge frequencies are identified by employing Improved Multisynchrosqueezing Transform (IMSST).

7. A method according to claim 1, wherein the step of processing the first set (250) of vibration response data and the second set (260) of vibration response data comprises employing at least one data processing technique that uses the first set of vibration response data as baseline data and the second set of vibration response data as test data, to identify a structural change.

8. A method according to claim 7, wherein the at least one data processing technique comprises at least one of:a physics-guided time-domain model configured to analyze time-domain signals in the first set (250) of vibration response data and the second set (260) of vibration response data;a feature extraction technique to extract change-sensitive features from the first set of vibration response data and the second set of vibrationresponse data, and a classification technique (830) to classify a bridge health state (850) based on the change-sensitive features;a deep auto-encoder (DAE) (630) configured to detect the structural change to the bridge based on a comparison of short-time data between the first set of vibration response data and the second set of vibration response data.

9. A method according to claim 8, wherein the feature extraction technique is Mel-Frequency Cepstral Coefficients (MFCCs)-based feature extraction.

10. A method according to claim 8 or 9, wherein the classification technique (830) is Optimized AdaBoost-linear support vector machine (840).

11. A method according to any of the preceding claims, wherein the at least one structural characteristic comprises at least one structural change, the method further comprising:calculating an identified damage ratio (IDR.) as a ratio of a number of detections of the at least one structural change of the bridge (102, 504, 608) to a total number of traversals; andestimating a damage severity, based on the identified damage ratio.

12. A method according to any of the preceding claims, further comprising:generating an alert signal (290) associated with identification of the at least one structural characteristic of the bridge (102, 504, 608); and sending the alert signal to a user device (292).

13. A method according to any of the preceding claims, further comprising:collecting sensor data (272, 282) from at least one third sensor (270, 280) arranged in each vehicle amongst the first set (230) of vehicles and the second set (240) of vehicles;detect when a given vehicle amongst the first set of vehicles and the second set of vehicles traverses over the bridge (102, 504, 608), by processing the sensor data; andactivating the at least one first sensor (210) for collecting the first set (250) of vibration response data and activating the at least one second sensor (220) for collecting the second set (260) of vibration response data, only while it is detected that the given vehicle traverses over the bridge.

14. A system (200) for identifying at least one structural characteristic of a bridge (102, 504, 608), the system comprising at least one processor (104, 204) communicably coupled to at least one first sensor (106, 210) and at least one second sensor (220), wherein the at least one processor is configured to:receive, from the at least one first sensor, a first set (250) of vibration response data associated with a first plurality of traversals of a first set of vehicles over the bridge, wherein the at least one first sensor is arranged in each vehicle amongst the first set of vehicles;receive, from the at least one second sensor, a second set (260) of vibration response data associated with a second plurality of traversals of a second set of vehicles over the bridge, wherein the at least one second sensor is arranged in each vehicle amongst the second set of vehicles; andprocess the first set of vibration response data and the second set of vibration response data, to identify the at least one structural characteristic.

15. A system according to claim 14, further comprising the at least one first sensor (106, 210) configured to measure the first set (250) of vibration response data, and the at least one second sensor (220) configured to measure the second set (260) of vibration response data.

16. A system according to claim 14 or 15, wherein the at least one first sensor (106, 210) is arranged in an axle that couples a pair of wheels of each vehicle amongst the first set (230) of vehicles, and wherein the at least one second sensor (220) is arranged in an axle that couples a pair of wheels of each vehicle amongst the second set (240) of vehicles.

17. A system according to any claims 14-16, wherein the at least one first sensor (106, 210) and / or the at least one second sensor (220) comprises at least one of: an accelerometer (212, 222), a motion sensor, a vibration sensor (224), a displacement sensor.

Images