This paper intends to present a synthesis of works based on the study of the behavior of the Kalman filters in two different domains. The first area is dedicated to the flocculation process occurring in water treatment. The second one covers aircraft structural damage based on SHM approach. The general methodology consists in modeling a state-parameter observer to perform estimations using the joint EKF. The robustness and efficiency of the Kalman filters is addressed in order to allow a cross-fertilization in the two area. Model propagation and accurate prognostics are therefore allowed due to the improvement of model parameter knowledge. In the context of the flocculation, the prediction of the time evolution of the characteristic diameters is much more efficient than QMOM. For fatigue damage prognostic, the best initial conditions leading to accurate estimation are highlighted according to materials. Whatever the problem is, the estimation error magnitude is known.
{"title":"A STATE-PARAMETER ESTIMATION IN TWO ENGINEERING DOMAINS: AN EXTENDED KALMAN FILTER APPROACH","authors":"L. Cot, S. Déjean, Carole Saudejaud","doi":"10.12783/shm2021/36286","DOIUrl":"https://doi.org/10.12783/shm2021/36286","url":null,"abstract":"This paper intends to present a synthesis of works based on the study of the behavior of the Kalman filters in two different domains. The first area is dedicated to the flocculation process occurring in water treatment. The second one covers aircraft structural damage based on SHM approach. The general methodology consists in modeling a state-parameter observer to perform estimations using the joint EKF. The robustness and efficiency of the Kalman filters is addressed in order to allow a cross-fertilization in the two area. Model propagation and accurate prognostics are therefore allowed due to the improvement of model parameter knowledge. In the context of the flocculation, the prediction of the time evolution of the characteristic diameters is much more efficient than QMOM. For fatigue damage prognostic, the best initial conditions leading to accurate estimation are highlighted according to materials. Whatever the problem is, the estimation error magnitude is known.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115393488","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Induction motors are one of the major electrical prime movers in industrial sectors. Since these motors are operated continuously, they are subjected to wear and tear which lead to faults at a later stage in its life. These faults which arise can be classified into 5 major categories i.e., broken rotor bars, stator winding faults, air-gap eccentricity, bearing faults, and torque fluctuations. A failure in induction motors leads to machine downtime, increased maintenance costs, and puts the lives of the plant personnel at risk, thus leading to undesirable consequences. Hence, uninterrupted operation of the machine is the need of the hour for which real-time condition-based monitoring of induction motors needs to be implemented. Industries are making an attempt to tap into the technology that involves around cyber-physical systems (CPS) and access real-time information regarding the motor health condition. The present article explores the CPS structure for real-time fault identification so that appropriate action can be taken by plant personnel. The CPS technology is a modular framework, which consists of a current sensor that transmits data to a remote minicomputer (e.g., Intel NUC kit) or a microcontroller (e.g., Raspberry Pi) by processing it through a data acquisition (DAQ) system across a wireless network. Since the range of defect frequencies for fault diagnosis in these induction motors is 5 kHz, Nyquist sampling frequency (𝐹𝑠) for data acquisition should at least be 10 kHz. It is to be noted that a microcontroller can be of low cost; however, maintaining 𝐹𝑠 more than 500 Hz tends to cause random jitters at the core of the operating system (OS). As a result, the signal-to-noise ratio (SNR) is compromised in microcontrollers leading to incorrect post-processing of the current time-stamp data for motor fault diagnosis. Hence, in the present article, a minicomputer is used for data acquisition of current time data at 𝐹𝑠 of 10 kHz and infer the motor health status by investigating the current spectrum. The information of motor health condition is stored in comma-separated values (CSV) file, which is further transferred over Google Cloud Storage (GCS) via hypertext transfer protocol (HTTP) with transport-layer security (TLS) encryption. HTTP converts the CSV data file into binary format and maintains the record of meta-data of the files. Meta-data essentially keeps track of when the file was created in the remote minicomputer. Additionally, in order to ensure a high data transfer rate at a given instant of time, the HTTP file transfer protocol divides the actual data into small chunks that are subjected to parallel composite uploads. When the data is collected in the computer at the receiver’s end i.e., the plant personnel in the present case, the data is recreated back to the original CSV file. As a result, the concerned plant personnel has complete information about the specific motor which has started failing and prevents any major break
{"title":"A CYBER-PHYSICAL SYSTEM BASED REAL-TIME FAULT DIAGNOSIS OF INDUCTION MOTORS","authors":"A. Mohanty, R. Pal","doi":"10.12783/shm2021/36275","DOIUrl":"https://doi.org/10.12783/shm2021/36275","url":null,"abstract":"Induction motors are one of the major electrical prime movers in industrial sectors. Since these motors are operated continuously, they are subjected to wear and tear which lead to faults at a later stage in its life. These faults which arise can be classified into 5 major categories i.e., broken rotor bars, stator winding faults, air-gap eccentricity, bearing faults, and torque fluctuations. A failure in induction motors leads to machine downtime, increased maintenance costs, and puts the lives of the plant personnel at risk, thus leading to undesirable consequences. Hence, uninterrupted operation of the machine is the need of the hour for which real-time condition-based monitoring of induction motors needs to be implemented. Industries are making an attempt to tap into the technology that involves around cyber-physical systems (CPS) and access real-time information regarding the motor health condition. The present article explores the CPS structure for real-time fault identification so that appropriate action can be taken by plant personnel. The CPS technology is a modular framework, which consists of a current sensor that transmits data to a remote minicomputer (e.g., Intel NUC kit) or a microcontroller (e.g., Raspberry Pi) by processing it through a data acquisition (DAQ) system across a wireless network. Since the range of defect frequencies for fault diagnosis in these induction motors is 5 kHz, Nyquist sampling frequency (𝐹𝑠) for data acquisition should at least be 10 kHz. It is to be noted that a microcontroller can be of low cost; however, maintaining 𝐹𝑠 more than 500 Hz tends to cause random jitters at the core of the operating system (OS). As a result, the signal-to-noise ratio (SNR) is compromised in microcontrollers leading to incorrect post-processing of the current time-stamp data for motor fault diagnosis. Hence, in the present article, a minicomputer is used for data acquisition of current time data at 𝐹𝑠 of 10 kHz and infer the motor health status by investigating the current spectrum. The information of motor health condition is stored in comma-separated values (CSV) file, which is further transferred over Google Cloud Storage (GCS) via hypertext transfer protocol (HTTP) with transport-layer security (TLS) encryption. HTTP converts the CSV data file into binary format and maintains the record of meta-data of the files. Meta-data essentially keeps track of when the file was created in the remote minicomputer. Additionally, in order to ensure a high data transfer rate at a given instant of time, the HTTP file transfer protocol divides the actual data into small chunks that are subjected to parallel composite uploads. When the data is collected in the computer at the receiver’s end i.e., the plant personnel in the present case, the data is recreated back to the original CSV file. As a result, the concerned plant personnel has complete information about the specific motor which has started failing and prevents any major break","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125212999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
C. Wickramarachchi, Xiaofei Jiang, E. Cross, K. Worden
Data-based SHM is highly dependent on the quality of the training data needed for machine learning algorithms. In many cases of engineering interest, data can be scarce, and this is a problem. However, in some cases, data are abundant and can create a computational burden. In data-rich situations, it is often desirable to select the subset(s) of the data which are of highest value (in some sense) for the problem of interest. In this paper, ‘value’ is interpreted in terms of information content, and entropy is used a measure of that content in order to condense training data without compromising useful information. Using the minimum covariance determinant, the dataset is first separated using inclusive outlier analysis. The entropies of the separated datasets are then assessed using parametric and nonparametric density estimators to identify the subset of data carrying most information. The Z24-Bridge dataset is used here to illustrate the idea, where the entropy values indicate that the subset containing data from environmental variations and damage is most rich in information. This subset was made up of half of the entire dataset, suggesting that it is possible to significantly reduce the amount of training data for an SHM algorithm whilst retaining the required information for analysis.
{"title":"ASSESSING THE INFORMATION CONTENT OF DATASETS FOR STRUCTURAL HEALTH MONITORING","authors":"C. Wickramarachchi, Xiaofei Jiang, E. Cross, K. Worden","doi":"10.12783/shm2021/36355","DOIUrl":"https://doi.org/10.12783/shm2021/36355","url":null,"abstract":"Data-based SHM is highly dependent on the quality of the training data needed for machine learning algorithms. In many cases of engineering interest, data can be scarce, and this is a problem. However, in some cases, data are abundant and can create a computational burden. In data-rich situations, it is often desirable to select the subset(s) of the data which are of highest value (in some sense) for the problem of interest. In this paper, ‘value’ is interpreted in terms of information content, and entropy is used a measure of that content in order to condense training data without compromising useful information. Using the minimum covariance determinant, the dataset is first separated using inclusive outlier analysis. The entropies of the separated datasets are then assessed using parametric and nonparametric density estimators to identify the subset of data carrying most information. The Z24-Bridge dataset is used here to illustrate the idea, where the entropy values indicate that the subset containing data from environmental variations and damage is most rich in information. This subset was made up of half of the entire dataset, suggesting that it is possible to significantly reduce the amount of training data for an SHM algorithm whilst retaining the required information for analysis.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116784241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yuning Wu, Chi-Luen Huang, Sangmin Lee, Keping Zhang, J. Popovics, M. Dersch, Xuan Zhu
With increasingly frequent extreme heat events over the past half century, thermal stress measurement and management of continuous welded rail (CWR) have become more important for railroad maintenance. Methods, including visual inspections and rail lifting, are routinely performed in railroad networks of the U.S. to prevent rail thermal buckling. When intervention becomes necessary, a rail distressing process, involving rail cutting and welding, will be performed to re-establish the zero-stress state at a desirable temperature. And the temperature at which the rail is stress-free is defined as rail neutral temperature (RNT). In this work, an RNT predictive tool that exploits zero group velocity (ZGV) modes and machine learning is proposed. First, the existence of ZGV modes in CWR is investigated through numerical simulation, using both semianalytical finite element analysis (SAFE) and finite element (FE) models. Further, parametric studies are performed to quantify the effect of axial loads and rail temperature on ZGV modes. Additionally, the team established an instrumented field test site at a revenue-service line and performed multi-day data collection to cover a wide range of temperature and thermal stress levels. FE models were calibrated based on the field-collected vibrational data via a linear program optimization approach and an excellent agreement between model and experimental results was obtained. Finally, a supervised learning framework was developed to estimate the RNT using rail temperature and resonance frequencies as the inputs. The results show that the proposed framework can provide RNT estimation with reasonable accuracy (±5 ºF) when measurement noise is low.
{"title":"RAIL NEUTRAL TEMPERATURE ESTIMATION USING ZERO GROUP VELOCITY MODES AND MACHINE LEARNING","authors":"Yuning Wu, Chi-Luen Huang, Sangmin Lee, Keping Zhang, J. Popovics, M. Dersch, Xuan Zhu","doi":"10.12783/shm2021/36296","DOIUrl":"https://doi.org/10.12783/shm2021/36296","url":null,"abstract":"With increasingly frequent extreme heat events over the past half century, thermal stress measurement and management of continuous welded rail (CWR) have become more important for railroad maintenance. Methods, including visual inspections and rail lifting, are routinely performed in railroad networks of the U.S. to prevent rail thermal buckling. When intervention becomes necessary, a rail distressing process, involving rail cutting and welding, will be performed to re-establish the zero-stress state at a desirable temperature. And the temperature at which the rail is stress-free is defined as rail neutral temperature (RNT). In this work, an RNT predictive tool that exploits zero group velocity (ZGV) modes and machine learning is proposed. First, the existence of ZGV modes in CWR is investigated through numerical simulation, using both semianalytical finite element analysis (SAFE) and finite element (FE) models. Further, parametric studies are performed to quantify the effect of axial loads and rail temperature on ZGV modes. Additionally, the team established an instrumented field test site at a revenue-service line and performed multi-day data collection to cover a wide range of temperature and thermal stress levels. FE models were calibrated based on the field-collected vibrational data via a linear program optimization approach and an excellent agreement between model and experimental results was obtained. Finally, a supervised learning framework was developed to estimate the RNT using rail temperature and resonance frequencies as the inputs. The results show that the proposed framework can provide RNT estimation with reasonable accuracy (±5 ºF) when measurement noise is low.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126143456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Li Ai, Elhussien Elbatanouny, L. K C, M. Bayat, V. Soltangharaei, Michel Van Torren, P. Ziehl
Impact damage is one of the major threats to the integrity of aircraft control surfaces such as wings and elevators. The conventional and widely applied inspection approach is visual inspection which is time-consuming and subject to human error. The innovation of this paper lies in developing a smart sensing system by leveraging acoustic emission (AE) for the real-time detection and evaluation of impact damage on aircraft elevators. The challenge of this system is to deploy a minimal number of AE sensors on the aircraft due to the environmental restriction during the operation of the aircraft while still effectively evaluate the impact damage. A convolutional neural network (CNN) is employed to localize the impact and evaluate the damage level by analyzing the wavelet of signals obtained by a single AE sensor. The proposed approach is verified by an impact test carried out on a thermoplastic aircraft elevator. The results demonstrate the efficacy and potential of the proposed approach.
{"title":"DETECTION AND EVALUATION OF IMPACT DAMAGE ON AIRCRAFT CONTROL SURFACE USING ACOUSTIC EMISSION AND CONVOLUTION NEURAL NETWORK","authors":"Li Ai, Elhussien Elbatanouny, L. K C, M. Bayat, V. Soltangharaei, Michel Van Torren, P. Ziehl","doi":"10.12783/shm2021/36365","DOIUrl":"https://doi.org/10.12783/shm2021/36365","url":null,"abstract":"Impact damage is one of the major threats to the integrity of aircraft control surfaces such as wings and elevators. The conventional and widely applied inspection approach is visual inspection which is time-consuming and subject to human error. The innovation of this paper lies in developing a smart sensing system by leveraging acoustic emission (AE) for the real-time detection and evaluation of impact damage on aircraft elevators. The challenge of this system is to deploy a minimal number of AE sensors on the aircraft due to the environmental restriction during the operation of the aircraft while still effectively evaluate the impact damage. A convolutional neural network (CNN) is employed to localize the impact and evaluate the damage level by analyzing the wavelet of signals obtained by a single AE sensor. The proposed approach is verified by an impact test carried out on a thermoplastic aircraft elevator. The results demonstrate the efficacy and potential of the proposed approach.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116766871","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Monitoring bridge health using vibrations of drive-by vehicles has various benefits, such as no need for directly installing and maintaining sensors on the bridge. However, many of the existing drive-by monitoring approaches are based on supervised learning models that require labeled data from every bridge of interest, which is expensive and time-consuming, if not impossible, to obtain. This labeling requirement is further exacerbated by having multiple diagnostic tasks (e.g., damage detection, localization, and quantification) because they have different learning difficulties. To this end, we introduce a multi-task domain adaptation framework that transfers the damage diagnosis model learned from one bridge to a new bridge without requiring any labels from the new bridge in any of the tasks. Our framework trains a hierarchical neural network model in an adversarial way to extract task-shared and task-specific features that are informative to multiple diagnostic tasks and invariant across multiple bridges. We evaluate our framework on experimental laboratory data collected from 2 bridges and 3 vehicles. We achieve accuracies of 95% for damage detection, 93% for localization, and up to 72% for quantification, which are ~2 times improvements from a baseline method.
{"title":"A HIERARCHICAL DOMAIN-ADVERSARIAL AND MULTI-TASK LEARNING ALGORITHM FOR BRIDGE DAMAGE DIAGNOSIS USING A DRIVE-BY VEHICLE","authors":"Jingxiao Liu, Susu Xu, M. Berges, H. Noh","doi":"10.12783/shm2021/36277","DOIUrl":"https://doi.org/10.12783/shm2021/36277","url":null,"abstract":"Monitoring bridge health using vibrations of drive-by vehicles has various benefits, such as no need for directly installing and maintaining sensors on the bridge. However, many of the existing drive-by monitoring approaches are based on supervised learning models that require labeled data from every bridge of interest, which is expensive and time-consuming, if not impossible, to obtain. This labeling requirement is further exacerbated by having multiple diagnostic tasks (e.g., damage detection, localization, and quantification) because they have different learning difficulties. To this end, we introduce a multi-task domain adaptation framework that transfers the damage diagnosis model learned from one bridge to a new bridge without requiring any labels from the new bridge in any of the tasks. Our framework trains a hierarchical neural network model in an adversarial way to extract task-shared and task-specific features that are informative to multiple diagnostic tasks and invariant across multiple bridges. We evaluate our framework on experimental laboratory data collected from 2 bridges and 3 vehicles. We achieve accuracies of 95% for damage detection, 93% for localization, and up to 72% for quantification, which are ~2 times improvements from a baseline method.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124965144","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Soheil Sadeghi Eshkevari, S. S. Eshkevari, Debarshi Sen, S. Pakzad
To maintain structural integrity and functionality structures are designed to accommodate operational loads as well as natural hazards during their lifetime. Active control systems are an efficient solution for structural response control when a structure is subjected to unexpected extreme loads. However, development of these systems through traditional means is limited by their model dependent nature. Recent advancements in adaptive learning methods, in particular, reinforcement learning (RL), for real-time decision-making problems, along with rapid growth in high-performance computational resources, enable structural engineers to transform the classic modelbased active control problem to a purely data-driven one. In this paper, we present a novel RL-based approach for designing active controllers by introducing RL-Controller, a flexible and scalable simulation environment. RL-Controller includes attributes and functionalities that are necessary to model active structural control mechanisms in detail. We show that the proposed framework is easily trainable for a five-story benchmark linear building with 65% reductions on average in inter story drifts (ISD) when subjected to strong ground motions. In a comparative study with an LQG active controller, we demonstrate that the proposed model-free algorithm learns actuator forcing strategies that yield higher performance, e.g., 25% more ISD reductions on average with respect to LQG, without using prior information about the mechanical properties of the system.
{"title":"STRUCTURAL ACTIVE CONTROL FRAMEWORK USING REINFORCEMENT LEARNING","authors":"Soheil Sadeghi Eshkevari, S. S. Eshkevari, Debarshi Sen, S. Pakzad","doi":"10.12783/shm2021/36293","DOIUrl":"https://doi.org/10.12783/shm2021/36293","url":null,"abstract":"To maintain structural integrity and functionality structures are designed to accommodate operational loads as well as natural hazards during their lifetime. Active control systems are an efficient solution for structural response control when a structure is subjected to unexpected extreme loads. However, development of these systems through traditional means is limited by their model dependent nature. Recent advancements in adaptive learning methods, in particular, reinforcement learning (RL), for real-time decision-making problems, along with rapid growth in high-performance computational resources, enable structural engineers to transform the classic modelbased active control problem to a purely data-driven one. In this paper, we present a novel RL-based approach for designing active controllers by introducing RL-Controller, a flexible and scalable simulation environment. RL-Controller includes attributes and functionalities that are necessary to model active structural control mechanisms in detail. We show that the proposed framework is easily trainable for a five-story benchmark linear building with 65% reductions on average in inter story drifts (ISD) when subjected to strong ground motions. In a comparative study with an LQG active controller, we demonstrate that the proposed model-free algorithm learns actuator forcing strategies that yield higher performance, e.g., 25% more ISD reductions on average with respect to LQG, without using prior information about the mechanical properties of the system.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130528668","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ultrasonic nondestructive testing is a promising method for performing damage assessments on concrete subjected to alkali-silica reactions (ASRs). Previous research incorporated only some ultrasonic wave parameters, and the other information from the ultrasonic signals was discarded. In this work, 13 features, including wave velocity and wavelet features, were extracted from the ultrasonic signals. A curve-fitting method was used to fit a polynomial relationship between the wave velocity and expansion of one concrete sample subjected to ASR to predict the expansion of another concrete sample subjected to ASR. Support vector regression (SVR), a machine learning model, was trained using all 13 features derived from the ultrasonic data obtained from the ASR samples. The SVR was then tested using the datasets from the ASR-2D sample. The performance showed that the curve-fitting method and the SVR had poor prediction results on the expansion of the ASR-2D sample. With feature selection, the performance of the SVR model using six selected features was significantly improved.
{"title":"MACHINE LEARNING OF ULTRASONIC DATA FOR EXPANSION PREDICTION OF CONCRETE WITH ALKALI-SILICA REACTION","authors":"Hongbin Sun, Jinying Zhu, P. Ramuhalli","doi":"10.12783/shm2021/36321","DOIUrl":"https://doi.org/10.12783/shm2021/36321","url":null,"abstract":"Ultrasonic nondestructive testing is a promising method for performing damage assessments on concrete subjected to alkali-silica reactions (ASRs). Previous research incorporated only some ultrasonic wave parameters, and the other information from the ultrasonic signals was discarded. In this work, 13 features, including wave velocity and wavelet features, were extracted from the ultrasonic signals. A curve-fitting method was used to fit a polynomial relationship between the wave velocity and expansion of one concrete sample subjected to ASR to predict the expansion of another concrete sample subjected to ASR. Support vector regression (SVR), a machine learning model, was trained using all 13 features derived from the ultrasonic data obtained from the ASR samples. The SVR was then tested using the datasets from the ASR-2D sample. The performance showed that the curve-fitting method and the SVR had poor prediction results on the expansion of the ASR-2D sample. With feature selection, the performance of the SVR model using six selected features was significantly improved.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123209440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ian T. Cummings, Elena C. Reinisch, Erica M. Jacobson, David H. Fraser, A. Wachtor, Eric B. Flynn
Acoustic Steady State Excitation Spatial Spectroscopy (ASSESS) is an ultrasonic inspection technique that was developed to rapidly evaluate large structures and identify regions of damage. An ultrasonic transducer affixed to the structure emits a single tone, and a scanning laser Doppler vibrometer (LDV) records the structure’s steady-state surface velocity response. Previous work has shown how local wavenumber can be estimated from the complex steady-state velocity response. This process has proved successful in detecting corrosion defects, delaminations, and regions of varying thickness. This work introduces a new processing method that utilizes a LiDAR generated point cloud representation of the scan region to identify and extract large planar sections from the measurement without causing distortion in the final wavenumber estimates. This new method uses the RANSAC algorithm to robustly extract planar sections and maps the complex steady-state response data onto a uniform grid on the detected planes. This is done in order to facilitate the use of an existing wavenumber estimation technique. We present wavefield and wavenumber results generated by applying this algorithm on a real-world dataset from a large area scan in an industrial structure with steel walls containing stringers, columns, and regions with different thicknesses.
{"title":"USING LIDAR TO IDENTIFY PLANAR MEASUREMENT REGIONS IN ULTRASONIC INSPECTIONS OF COMPLEX STRUCTURES","authors":"Ian T. Cummings, Elena C. Reinisch, Erica M. Jacobson, David H. Fraser, A. Wachtor, Eric B. Flynn","doi":"10.12783/shm2021/36318","DOIUrl":"https://doi.org/10.12783/shm2021/36318","url":null,"abstract":"Acoustic Steady State Excitation Spatial Spectroscopy (ASSESS) is an ultrasonic inspection technique that was developed to rapidly evaluate large structures and identify regions of damage. An ultrasonic transducer affixed to the structure emits a single tone, and a scanning laser Doppler vibrometer (LDV) records the structure’s steady-state surface velocity response. Previous work has shown how local wavenumber can be estimated from the complex steady-state velocity response. This process has proved successful in detecting corrosion defects, delaminations, and regions of varying thickness. This work introduces a new processing method that utilizes a LiDAR generated point cloud representation of the scan region to identify and extract large planar sections from the measurement without causing distortion in the final wavenumber estimates. This new method uses the RANSAC algorithm to robustly extract planar sections and maps the complex steady-state response data onto a uniform grid on the detected planes. This is done in order to facilitate the use of an existing wavenumber estimation technique. We present wavefield and wavenumber results generated by applying this algorithm on a real-world dataset from a large area scan in an industrial structure with steel walls containing stringers, columns, and regions with different thicknesses.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121422624","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A novel model updating-based damage detection method is proposed that uses the Unwrapped Instantaneous Hilbert Phase (UIHP) of the condensed frequency response function (CFRF) as input to the objective function of an optimisation problem. The novelty of the proposed method lies in two items: (1) using the CFRF instead of the FRF itself, and (2) using the UIHP associated with the columns of the CFRF as input. The proposed modifications bring about the following improvements in the damage detection practice as follows: (1) CFRF will reduce the number of required degrees of freedom (DOFs) to be measured, and (2) the UIHP mitigates the effect of the measurement noise on damage detection. The problem of damage detection in a laminated composite plate with different number of layers and ply orientation has been solved where the results demonstrate the effectiveness of the proposed method.
{"title":"MINIMIZING NOISE EFFECTS IN STRUCTURAL HEALTH MONITORING USING HILBERT TRANSFORM OF THE CONDENSED FRF","authors":"Sahar Hassani, M. Mousavi, A. Gandomi","doi":"10.12783/shm2021/36343","DOIUrl":"https://doi.org/10.12783/shm2021/36343","url":null,"abstract":"A novel model updating-based damage detection method is proposed that uses the Unwrapped Instantaneous Hilbert Phase (UIHP) of the condensed frequency response function (CFRF) as input to the objective function of an optimisation problem. The novelty of the proposed method lies in two items: (1) using the CFRF instead of the FRF itself, and (2) using the UIHP associated with the columns of the CFRF as input. The proposed modifications bring about the following improvements in the damage detection practice as follows: (1) CFRF will reduce the number of required degrees of freedom (DOFs) to be measured, and (2) the UIHP mitigates the effect of the measurement noise on damage detection. The problem of damage detection in a laminated composite plate with different number of layers and ply orientation has been solved where the results demonstrate the effectiveness of the proposed method.","PeriodicalId":180083,"journal":{"name":"Proceedings of the 13th International Workshop on Structural Health Monitoring","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125780404","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：