Data-driven Structural Health Monitoring Research Articles

Efficient layerwise multiphysics spectral element model for delaminated composite strips with PWAS transducers

Efficient structural element-based models are essential for fast simulations of wave propagation in composite structures for model-based and physics-informed data-driven structural health monitoring. This article introduces the first efficient multiphysics time-domain spectral structural element for wave propagation analysis of beam and panel-type composite structures with piezoelectric transducers (patch or full layers) containing delaminations. A general framework is presented to model multiple delaminations and transducer patches located arbitrarily. The intact host and patch transducer-bonded laminates and sub-laminates between delaminations are modelled separately using an electromechanically coupled efficient layerwise zigzag theory (ZIGT) for kinematics and a piecewise quadratic variation for the electric potential across piezoelectric layers. The high-order spectral element (SE) features a virtual electric node to model equipotential surfaces of piezoelectric transducers apart from the usual physical nodes having mechanical and internal electric degrees of freedom. A hybrid point-least squares continuity approach is employed to maintain continuity at the intersections of delaminated sub-laminates or the patch-bonded laminate with the host laminate. The model’s performance in capturing electroelastic waves’ interaction with delamination is examined with reference to the conventional finite element (FE) solution based on the ZIGT, continuum-based FE solutions, and the SE solution based on a non-layerwise version of the laminate theory. Finally, the model is used to examine the impact of interfacial location and size of delaminations on wave propagation behaviour.

Data-Driven Multi-Fault Detection in Pipelines Utilizing Frequency Response Function and Artificial Neural Networks

This research presents a data-driven structural health monitoring (SHM) approach for pipeline systems that leverages frequency response function (FRF) signals and artificial neural network (ANN) algorithms to accurately identify and classify diverse pipeline fault conditions. The study focuses on three specific faults: bolt looseness, scale deposits, and crack occurrence at pipeline supports, which were replicated on a pipeline segment located at the Sound and Vibration Research Group (SVRG) at University Putra Malaysia (UPM). The FRF signals were captured using accelerometers to monitor the structural health of the pipeline. The data acquisition stage involved collecting FRF signals from the accelerometers to capture vibrations and responses related to the identified faults using a Siemens LMS SCADAS data acquisition unit. The data underwent preprocessing, including the application of principal component analysis (PCA) for feature selection. The subsequent data processing stage involved the application of an artificial neural network (ANN) algorithm for pattern recognition to analyze and classify the acquired data, identifying patterns associated with the replicated fault conditions. The proposed methodology demonstrated exceptional performance, with the ANN model achieving consistently high overall accuracy (above 99.7%) and remarkably low mean squared error (in the range of 0.0088×10−3to0.3062×10−3) across multiple iterations and sensor datasets. The detailed class-specific metrics, including accuracy, precision, sensitivity, and F1-score, further substantiated the model's effectiveness in identifying the individual fault types with near-perfect or perfect results for the majority of the fault scenarios. The location-invariant performance of the ANN model across different sensor placements demonstrates the robustness of the proposed data-driven SHM methodology. This research highlights the transformative potential of integrating state-of-the-art data-driven techniques to revolutionize the monitoring and assessment of critical pipeline infrastructure, ultimately enhancing the safety, reliability, and longevity of these vital systems.

An unsupervised damage detection method for an operating wind turbine blade

Wind turbine blades are normally exposed to significant dynamic loads over their lifetime, including extreme conditions. This can lead to several types of damages such as cracking and delamination. Monitoring the structural condition of the blades to detect damages before they become catastrophic can be important. In recent years, the concept of data-driven structural health monitoring (SHM) approaches for wind turbine blade damage detection has become popular. These methods employ features collected from time series obtained from monitored wind turbine blades to identify damages. However, selecting an optimum number of features, which can reduce computational cost and minimize the effects of environmental and operational conditions, is still a challenge. In this paper, a data-driven approach is developed combining a feature extraction strategy and un-supervised anomaly detection. Monitoring data collected from several accelerometers mounted on a blade of an operating Vestas V27 wind turbine in healthy and damaged scenarios are used in this regard. Several time-domain, temporal, and time-frequency domain features are employed for the purpose of anomaly detection using unsupervised learning methods. Isolation Forest and One-class Support Vector Machine are used to train for anomaly detection. A threshold of detection is established to distinguish between the healthy and damaged states. Results show that the proposed approach can successfully identify the damage cases, but environmental and operational conditions still have some effect when it comes to identifying the extent of the damage.

Hierarchical verification and validation in a forward model-driven structural health monitoring strategy

This paper presents a demonstrative application of a forward model-driven approach to structural health monitoring (SHM), incorporating hierarchical validation methods. A key tenet of the approach is that an SHM system can be constructed that is capable of diagnosing damage at the full system level, without full system damage-state data having been used in its development; achieving this would be highly impactful as the system-level damage state data is generally not feasible to acquire (previous SHM methods such as data-driven SHM have been hindered by their dependence on these data). This is achieved by carrying out validation activities on the damage model at the subassembly level of the structure. The particular focus of the present paper is on damage detection and assessment, although the approach offers a natural basis for extension to other damage identification activities such as damage location and prognosis. The present study focuses on two of the key elements of the model-driven approach: validation of the predictive substructure models and their application in the assembled model. The ideas discussed are demonstrated in a case study based on a laboratory-scale truss bridge structure.

Data-Driven Monitoring and Predictive Maintenance for Engineering Structures: Technologies, Implementation Challenges, and Future Directions

Estimating engineering structures’ health conditions and predicting their future behaviors are fundamental problems for a city’s safe and efficient operations. Data-driven solutions estimate the health conditions using statistical models generated from measurement data. They have attracted growing interest recently because advances in information and communication technologies (ICT) have enabled numerous real-world measurement data, and the flourishing big data community has provided enormous state-of-the-art data analytics algorithms. Nevertheless, most existing studies remain in numerical simulation while neglecting their real-world implementation, restricting the extensive development of data-driven methods. In this survey, we provide a structural overview of the past decade’s data-driven structural health monitoring (SHM) systems and algorithms from the perspective of real-world implementation. Specifically, we cover various aspects of the design and implementation of monitoring systems, including sensing technologies, communication technologies, and processing software. In addition, we classify the used data sources and statistical models with fined details. Under the proposed taxonomy, their limitations and advantages are thoroughly discussed. Based on our insights into existing studies, we clarify two major implementation challenges: the digressive performance in the real-world environment and inefficient computing systems for real-time data analytics. Possible solutions are then proposed to mitigate those challenges and promote the implementation of data-driven methods. Finally, we raise our outlook on future trends and suggest promising directions for further investigation.

Unsupervised acoustic detection of fatigue-induced damage modes from wind turbine blades

This paper proposes a new in-situ damage detection approach for wind turbine blades, which leverages blade-internal non-stationary acoustic pressure fluctuations caused by the mechanical loading as the main source of excitation. This acoustic excitation was leveraged for the detection of fatigue-related damage modes on a full-scale wind turbine blade undergoing edgewise fatigue testing. An unsupervised, data-driven structural health monitoring strategy was developed to learn the normal cavity-internal acoustic sequences generated by the blade’s load cycles and to detect damage-related anomalies in the context of those sequences. A linear cepstral-coefficient based feature set was used to characterize the cavity-internal acoustics and LSTM-autoencoders were trained to accurately reconstruct healthy-case sequences. The reconstruction error was then used to characterize anomalous acoustic patterns within the blade cavity. The technique was able to detect a damage event earlier than a strain-based system by 120,000 load cycles.

Structural damage severity classification from time-frequency acceleration data using convolutional neural networks

Structural damage identification involves interpretation of vibration measurements to identify patterns that are indicative of changes in structural characteristics. Identification of such patterns using machine learning (ML) algorithms is a primary task in data-driven structural health monitoring. However, providing ML algorithms with features that are sensitive to damage has always been a challenge for accurate damage detection. In recent years, deep Convolutional Neural Networks (CNNs) proved to be versatile and powerful tools to identify the underlying patterns directly from low-level raw data. This study explores the applicability of CNNs to damage severity classification. A new format of input data for CNN damage identifiers is proposed based on time–frequency representations of acceleration data from multiple sensors. First, Continuous Wavelet Transformation (CWT) of acceleration response is employed to derive the time–frequency data, then, the results of transformations from multiple accelerometers are stored in separate channels creating a three-dimensional block of data. Such input data provide a CNN with structural information that expands into time, frequency, and spatial domains. To evaluate the accuracy of such a CNN-based classifier, a dataset comprised of free vibration response of a concrete beam under impact hammer tests is utilized. It is shown that a CNN trained with CWT data can successfully classify various severity of the damage, from minor to severe, with 100% accuracy. Also, this study examines the sensitivity of the CNN classifier to the number of training samples. It is demonstrated that the accuracy of such a network has a relatively lower sensitivity to the number of training samples and/or the number of measurement points to the degree that it maintains an F1-score above 0.813 when trained with data from one impact point and one sensor. On the other hand, for the case of a CNN classifier trained directly with time-series acceleration data the value of the F1-score drops to 0.590. Moreover, when compared with a multilayer perceptron trained with the extracted modal properties, it is illustrated that CNN classifiers have a better capability in capturing key features from the measurement, and thus produce more accurate predictions. Therefore, the combination of CNN classifiers with time–frequency data from multiple sensors results in damage identifiers that are less sensitive to sample size and more robust to changes in excitations and measurement points.

A Loewner-Based System Identification and Structural Health Monitoring Approach for Mechanical Systems

Data-driven structural health monitoring (SHM) requires precise estimates of the target system behaviour. In this sense, SHM by means of modal parameters is strictly linked to system identification (SI). However, existing frequency-domain SI techniques have several theoretical and practical drawbacks. This paper proposes using an input-output system identification technique based on rational interpolation, known as the Loewner framework (LF), to estimate the modal properties of mechanical systems. Pioneeringly, the Loewner framework mode shapes and natural frequencies estimated by LF are then applied as damage-sensitive features for damage detection. To assess its capability, the Loewner framework is validated on both numerical and experimental datasets and compared to established system identification techniques. Promising results are achieved in terms of accuracy and reliability.

Unsupervised Learning Methods for Data-Driven Vibration-Based Structural Health Monitoring: A Review

Structural damage detection using unsupervised learning methods has been a trending topic in the structural health monitoring (SHM) research community during the past decades. In the context of SHM, unsupervised learning methods rely only on data acquired from intact structures for training the statistical models. Consequently, they are often seen as more practical than their supervised counterpart in implementing an early-warning damage detection system in civil structures. In this article, we review publications on data-driven structural health monitoring from the last decade that relies on unsupervised learning methods with a focus on real-world application and practicality. Novelty detection using vibration data is by far the most common approach for unsupervised learning SHM and is, therefore, given more attention in this article. Following a brief introduction, we present the state-of-the-art studies in unsupervised-learning SHM, categorized by the types of used machine-learning methods. We then examine the benchmarks that are commonly used to validate unsupervised-learning SHM methods. We also discuss the main challenges and limitations in the existing literature that make it difficult to translate SHM methods from research to practical applications. Accordingly, we outline the current knowledge gaps and provide recommendations for future directions to assist researchers in developing more reliable SHM methods.

Parametric Damage Mechanics Empowering Structural Health Monitoring of 3D Woven Composites.

This paper presents a data-driven structural health monitoring (SHM) method by the use of so-called reduced-order models relying on an offline training/online use for unidirectional fiber and matrix failure detection in a 3D woven composite plate. During the offline phase (or learning) a dataset of possible damage localization, fiber and matrix failure ratios is generated through high-fidelity simulations (ABAQUS software). Then, a reduced model in a lower-dimensional approximation subspace based on the so-called sparse proper generalized decomposition (sPGD) is constructed. The parametrized approach of the sPGD method reduces the computational burden associated with a high-fidelity solver and allows a faster evaluation of all possible failure configurations. However, during the testing phase, it turns out that classical sPGD fails to capture the influence of the damage localization on the solution. To alleviate the just-referred difficulties, the present work proposes an adaptive sPGD. First, a change of variable is carried out to place all the damage areas on the same reference region, where an adapted interpolation can be done. During the online use, an optimization algorithm is employed with numerical experiments to evaluate the damage localization and damage ratio which allow us to define the health state of the structure.

Damage Sensitive Signals for the Assessment of the Conditions of Wind Turbine Rotor Blades Using Electromagnetic Waves

One of the most important renewable energy technologies used nowadays are wind power turbines. In this paper, we are interested in identifying the operating status of wind turbines, especially rotor blades, by means of multiphysical models. It is a state-of-the-art technology to test mechanical structures with ultrasonic-based methods. However, due to the density and the required high resolution, the testing is performed with high-frequency waves, which cannot penetrate the structure in depth. Therefore, there is a need to adopt techniques in the fields of multiphysical model-based inversion schemes or data-driven structural health monitoring. Before investing effort in the development of such approaches, further insights and approaches are necessary to make the techniques applicable to structures such as wind power plants (blades). Among the expected developments, further accelerations of the so-called “forward codes” for a more efficient implementation of the wave equation could be envisaged. Here, we employ electromagnetic waves for the early detection of cracks. Because in many practical situations, it is not possible to apply techniques from tomography (characterized by multiple sources and sensor pairs), we focus here on the question of whether the existence of cracks can be determined by using only one source for the sent waves.

Online Hybrid Learning Methods for Real-Time Structural Health Monitoring Using Remote Sensing and Small Displacement Data

Structural health monitoring (SHM) by using remote sensing and synthetic aperture radar (SAR) images is a promising approach to assessing the safety and the integrity of civil structures. Apart from this issue, artificial intelligence and machine learning have brought great opportunities to SHM by learning an automated computational model for damage detection. Accordingly, this article proposes online hybrid learning methods to firstly deal with some major challenges in data-driven SHM and secondly detect damage via small displacement data from SAR images in a real-time manner. The proposed methods contain three main parts: (i) data augmentation by Hamiltonian Monte Carlo and slice sampling for addressing the problem of small displacement data, (ii) data normalization by an online deep transfer learning algorithm for removing the effects of environmental and/or operational variability from augmented data, and (iii) feature classification via a scalar novelty score. The major contributions of this research include proposing two online hybrid unsupervised learning methods and providing effective frameworks for online damage detection. A small set of displacement samples extracted from SAR images of TerraSar-X regarding a long-term monitoring scheme of the Tadcaster Bridge in United Kingdom is applied to validate the proposed methods.

Automatic Crack Classification by Exploiting Statistical Event Descriptors for Deep Learning

In modern building infrastructures, the chance to devise adaptive and unsupervised data-driven structural health monitoring (SHM) systems is gaining in popularity. This is due to the large availability of big data from low-cost sensors with communication capabilities and advanced modeling tools such as deep learning. A promising method suitable for smart SHM is the analysis of acoustic emissions (AEs), i.e., ultrasonic waves generated by internal ruptures of the concrete when it is stressed. The advantage in respect to traditional ultrasonic measurement methods is the absence of the emitter and the suitability to implement continuous monitoring. The main purpose of this paper is to combine deep neural networks with bidirectional long short term memory and advanced statistical analysis involving instantaneous frequency and spectral kurtosis to develop an accurate classification tool for tensile, shear and mixed modes originated from AE events (cracks). We investigated effective event descriptors to capture the unique characteristics from the different types of modes. Tests on experimental results confirm that this method achieves promising classification among different crack events and can impact on the design of the future of SHM technologies. This approach is effective to classify incipient damages with 92% of accuracy, which is advantageous to plan maintenance.

Data Clustering based on Gaussian Mixture Model and Expectation- Maximization Algorithm for Data-driven Structural Health Monitoring System

Data groups generated by a system often inherit dynamics characteristics unique in data distribution parameters. A degradation in structural health can affect the dynamic behavior hence the probability distribution parameters. Based on the probabilistic and Expectation-Maximization (EM) algorithm, Gaussian Mixture Model (GMM), one can cluster data groups that may overlap with different data groups based on different orientations and shapes. This article explores GMM probabilistic model applied on vibration data set generated by aircraft wing box structure for Structural Health Monitoring (SHM) application. In the data processing stage, the high dimensional data is transformed into lower dimensions using Kernel Principal Component Analysis (KPCA). KPCA transforms the continuous signal into discrete data, allowing the ellipsoids' fitting (clusters) on the data spread. Based on the baseline data set (undamaged structural condition) and several components (loading class and damage class), the fitting is performed using GMM driven by EM. This paper shows that GMM-EM based data clustering model is an effective clustering probability model in fitting the data density in the presence of operational variations. It highlights clustering of reduced vibration data using KPCA in the interest of SHM based on the baseline’s initial parameters.

A BIM Based Approach for Structural Health Monitoring of Bridges

With the development of bridges, independent condition assessment of large-scale bridges has garnered significant attention over the past few decades. Data-driven structural health monitoring (SHM) techniques offer valuable information on the existing health of the structures, maintain safety, and uninterrupted use under varied operational conditions by undertaking timely risk and hazard mitigation. Traditional approaches, however, are not enough to monitor a large amount of SHM data and conduct systematic decision making for future maintenance. In this paper, a bridge health monitoring system is developed through the combination of building information modeling (BIM) and traditional bridge health monitoring that can organize and visualize a considerable amount of sensor data and subsequent structural health information over a prolonged period. The system can identify the structural damage by evaluating the data from sensors using Wigner-Ville distribution (WVD) in bilinear time-frequency analysis. A BIM-enabled platform is utilized to develop the proposed visualization tool for a long-span bridge and enable automated sensor data inventory into the BIM environment. The system has been tested for its robustness and functionality against the development requirements, and the results showed promising potential to support more effective bridge information management.

Transfer-learning guided Bayesian model updating for damage identification considering modeling uncertainty

Modeling uncertainty or modeling error has been widely recognized as one major challenge in structural model updating for structural identification and damage detection. It renders model updating inherently ineffective in converging to the real structural model because of the physical bias present in establishing the numerical model. This study aims to minimize the influence of modeling uncertainty during model updating so that the updated model can accurately indicate the damage state. To this end, this study proposes a methodology that applies pattern recognition methods to guide Bayesian model updating (BMU) and supervise the identification of structural damage. In detail, the transfer learning (TL) technique realized by domain adaptation is used to bridge the gap between the biased numerical model and the real structure and to guide the model updating process. Numerical and experimental studies have been implemented to validate the efficiency of domain adaptation in identifying the correct damage locations and the advantage of TL-guided BMU over the traditional method in identifying damage severities when modeling error exists. Moreover, this study proposes applying domain adaptation to bridge the gap between model-based and data-driven structural health monitoring (SHM) that are realized via model updating and pattern recognition, respectively. The proposed methodology is valuable and instructive for future work in this area.

Online unsupervised detection of structural changes using train–induced dynamic responses

This paper exploits unsupervised data-driven structural health monitoring (SHM) in order to propose a continuous online procedure for damage detection based on train-induced dynamic bridge responses, taking advantage of the large-magnitude loading for enhancing sensitivity to small-scale structural changes. While such large responses induced by trains might create more damage-sensitive information in the measured response, it also amplifies the effects on those measurements from the environment. Thus, one of the biggest contributions herein is a methodology that exploits the large bridge responses induced by train passage while rejecting the confounding influences of the environment in such a way that false positive detections are mitigated. Furthermore, this research work introduces an adaptable confidence decision threshold that further improves damage detection over time. To ensure an online continuous assessment, a hybrid combination of autoregressive exogenous input (ARX) models, principal components analysis (PCA), and clustering algorithms was sequentially applied to the monitoring data, in a moving window process. A comparison between the performance obtained from autoregressive (AR) and ARX models as feature extractors was conducted, and it was concluded that ARX models lead to increased sensitivity to damage due to their ability to capture cross information between the sensors. The PCA proved its importance and effectiveness in removing observable changes induced by variations in train speed or temperature without the need to measure them, and the clustering methods allowed for an automatic classification of the damage-sensitive features. Since it was not possible to introduce damage to the bridge, several structural conditions were simulated with a highly reliable digital twin of the Sado Bridge, tuned with experimental data acquired from a SHM system installed on site, in order to test and validate the efficiency of the proposed procedure. The strategy proved to be robust when detecting a comprehensive set of damage scenarios with a false detection incidence of 2%. Moreover, it showed sensitivity to smaller damage levels (earlier in life), even when it consists of small stiffness reductions that do not impair structural safety and are imperceptible in the original signals.

Data-Driven Structural Health Monitoring Using Feature Fusion and Hybrid Deep Learning

Smart structural health monitoring (SHM) for large-scale infrastructure is an intriguing subject for engineering communities thanks to its significant advantages such as timely damage detection, optimal maintenance strategy, and reduced resource requirement. Yet, it is a challenging topic as it requires handling a large amount of collected sensors data continuously, which is inevitably contaminated by random noises. Therefore, this study developed a practical end-to-end framework that makes use of physical features embedded in raw data and an elaborated hybrid deep learning model, namely 1-DCNN-LSTM, featuring two algorithms—convolutional neural network (CNN) and long-short term memory (LSTM). In order to extract relevant features from sensory data, the method combines various signal processing techniques such as the autoregressive model, discrete wavelet transform, and empirical mode decomposition. The hybrid deep learning 1-DCNN-LSTM is designed based on the CNN’s capacity of capturing local information and the LSTM network’s prominent ability to learn long-term dependencies. Through three case studies involving both experimental and synthetic data sets, it is demonstrated that the proposed approach achieves highly accurate damage detection, as accurate as the powerful 2-D CNN, but with a lower time and memory complexity, making it suitable for real-time SHM. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This article aims to develop a practical data-driven method for automatically monitoring the operational state of structures. In order to achieve consistently and highly accurate results in performing different tasks for diverse structures, we combine underlying features in both time and frequency domains extracted from measured signal vibration data. Three popular data featuring methods are combined to achieve the diversity gain which would not be possible with each individual method. As the vibration is usually measured by long time-series signals, the most efficient deep learning architecture for time-series signal, namely long-short term memory (LSTM), is considered for this work. Besides, each structure has its own dynamic properties, i.e., eigenfrequencies, around which the most relevant information is in the frequency domain, thus convolutional neural network specifically designed for capturing local information is used in combination with LSTM, forming a hybrid deep learning architecture. The applicability and effectiveness of the proposed approach are supported by three case studies with different types of structures, showing highly accurate damage detection with reduced resource requirements. These advantages can be valuable for developing a model for live monitoring of structural health in the future life-line infrastructures.

Early damage assessment in large-scale structures by innovative statistical pattern recognition methods based on time series modeling and novelty detection

Time series analysis and novelty detection are effective and promising methods for data-driven structural health monitoring (SHM) based on the statistical pattern recognition paradigm. However, processing substantially large volumes of vibration measurements may represent a serious limitation, especially for long-term SHM programs of large-scale civil structures. Moreover, shortcomings like the choice of an appropriate time series model in an automatic manner, the determination of optimal orders of the identified model and the classification of random high-dimensional features for damage detection, can strongly affect the performance of these approaches. This study is intended to propose statistical pattern recognition methods regarding time series modeling for feature extraction and novelty detection in feature classification in the presence of big data. These methods include an automatic model identification algorithm, an improved order determination approach and a hybrid distance-based novelty detection through a combination of Partition-based Kullback-Leibler divergence and Mahalanobis-squared distance. Experimental datasets relevant to a cable-stayed bridge are considered to validate the effectiveness of the proposed methods. Results demonstrate that: the AutoRegressive-AutoRegressive with eXogenous input (AR-ARX) model turns out to be the most suitable representation for feature extraction; the orders of this model are efficiently and automatically determined; the proposed novelty detection approach is highly successful in detecting damage, even in case of large volumes of random high-dimensional features.

A numerical study on multi‐site damage identification: A data‐driven method via constrained independent component analysis

This paper presents a solution to the multi-site structural damage identification problem using a data-driven method and constrained independent component analysis (cICA). While existing studies in this field presented encouraging results for single-site damage identification, limited research effort has been devoted to identifying multi-site damage due its complexity. Efficient features for single-site damage identification may lose their effectiveness when multi-site damage occurs. This paper extracts damage-sensitive features from the ICA outcome of the structural responses under certain excitations. The information on structural damage contained in the response is compacted into the mixing matrix by enforcing identical independent components to that of intact structures. Hence, the cICA can significantly reduce the feature dimension and preserve all the valuable information of damage. A case study indicates that the mixing matrix elements, when used as damage features, can distinguish multi-site damage cases from single-site damage cases and locate the single-site damage. Furthermore, the mixing matrix columns of multi-site damage cases exhibit distinct correlation with that of the corresponding single-site damage cases. As a result, the proposed method can progressively locate the structural damage. Moreover, the present study has the potential to identify multi-site damage identification in data-driven structural health monitoring without requiring multi-site damage data as a reference. This relieves the burden from data incompleteness when using data-based damage identification methods and pattern recognition.