A stochastic model based on Markov random field is proposed to model the spatial distribution of vehicle loads on longspan bridges. The bridge deck is divided into a finite set of discrete grid cells, each cell has two states according to whether the cell is occupied by the heavy vehicle load or not, then a four-neighbor lattice-structured undirected graphical model with each node corresponding to a cell state variable is proposed to model the location distribution of heavy vehicle loads on the bridge deck. The node potential is defined to quantitatively describe the randomness of node state, and the edge potential is defined to quantitatively describe the correlation of the connected node pair. The junction tree algorithm is employed to obtain the systematic solutions of inference problems of the graphical model. A marked random variable is assigned to each node to represent the amplitude of the total weight of vehicle applied on the corresponding cell of the bridge deck. The rationality of the model is validated by a Monte Carlo simulation of a learned model based on monitored data of a cable-stayed bridge.
This paper proposes a novel and effective method in the field of Non-Destructive Evaluation (NDE). Traditional ultrasonic computerized tomography (UCT) is a heavy task to detect the damages in the object for the numerous measuring times and the huge cost of manual labor. However, utilizing the method proposed in this paper can effectively overcome this great disadvantage, the essence of the application of Compressive Sampling(CS) in the detection of the object is to selectively choose a small quantity of measuring path in the huge number of total measurements. Due to the sparsity of damages in concrete structure, the usage of CS is available. Firstly, we divide the object entirely into numerous grids in order to image the internal situation of the object respectively. Secondly, a measurement matrix to massively decline the quantity of the measuring time should be computed. Thirdly, the travel time of each path we selected according to the matrix should be acquired, utilizing these travel time by adopting the l1-minimization program can we consequently obtained the slowness of the elements inside the object, thus reconstruct the internal situation of the object clearly and effectively. Furthermore, by applying this method we proposed in this paper into the simulation we can not only determine the damage location but also figure the size of it out. Because of the massive decline of the measuring times and accurate reconstruction, we substantiate CS method applied into the monitoring of concrete structure proves to be a shortcut in the field of NDE.
KEYWORDS: Sensor networks, Sensors, Signal attenuation, Bridges, Data transmission, Structural health monitoring, Compressed sensing, Algorithm development, Reconstruction algorithms, Signal processing
In wireless data transmissions processes, the problem of packet loss become an important factor that affects the
robustness of wireless data transmission. In order to solve the problem, the compressive sensing (CS) based wireless data
transmission approach is proposed in this paper. The specific steps that use the CS approach to reconstruct lost data are as
follows: The first step is to encode the original data in a random sampling matrix. Then the original data and the encoded
data are sent to the receiving side through the wireless transmission. After the data is received, if there are packets lost, it
can be reconstructed by CS approach. The reconstructed data are able to compensate for the incomplete original data in a
certain range. In this paper, a wireless sensor network (WSN) based on Wi-Fi is developed for verifying the effectiveness
and feasibility of the CS approach. The WSN consists of small nodes with sensors, base stations, PC client. Experimental
results show that the wireless sensor network is working properly and steady. Moreover, the CS approach could
compensate for the packet loss effectively, and increase the robustness and the speed of wireless transmission greatly.
Data loss is a common problem for monitoring systems based on wireless sensors. Reliable communication protocols, which enhance communication reliability by repetitively transmitting unreceived packets, is one approach to tackle the problem of data loss. An alternative approach allows data loss to some extent and seeks to recover the lost data from an algorithmic point of view. Compressive sensing (CS) provides such a data loss recovery technique. This technique can be embedded into smart wireless sensors and effectively increases wireless communication reliability without retransmitting the data. The basic idea of CS-based approach is that, instead of transmitting the raw signal acquired by the sensor, a transformed signal that is generated by projecting the raw signal onto a random matrix, is transmitted. Some data loss may occur during the transmission of this transformed signal. However, according to the theory of CS, the raw signal can be effectively reconstructed from the received incomplete transformed signal given that the raw signal is compressible in some basis and the data loss ratio is low. This CS-based technique is implemented into the Imote2 smart sensor platform using the foundation of Illinois Structural Health Monitoring Project (ISHMP) Service Tool-suite. To overcome the constraints of limited onboard resources of wireless sensor nodes, a method called random demodulator (RD) is employed to provide memory and power efficient construction of the random sampling matrix. Adaptation of RD sampling matrix is made to accommodate data loss in wireless transmission and meet the objectives of the data recovery. The embedded program is tested in a series of sensing and communication experiments. Examples and parametric study are presented to demonstrate the applicability of the embedded program as well as to show the efficacy of CS-based data loss recovery for real wireless SHM systems.
In the authors’ previous work, an inductive substructure identification method was proposed for shear structures, which utilizes the frequency responses (Fourier transforms) of floor accelerations to formulate a series of inductive substructure identification problems, estimating the structural parameters from top to bottom iteratively. However, the simulation results show that the proposed method can only obtain relatively accurate results if measurement noise is not large. In order to improve the identification accuracy, an uncertainty analysis of the parameter identification errors is conducted for this method in this paper, revealing the important factors that influence the identification accuracy. Based on this result, a new substructure identification method is proposed herein, in which the cross power spectral densities (CPSDs) of structural responses, computed via multi-taper method, are utilized to formulate the substructure identification problems. A similar uncertainty analysis of the identification errors is carried out for the new method, illustrating why the new method could significantly improve the identification accuracy. Finally, a numerical example of 8-story shear building structure is utilized to verify the effectiveness of the new multi-taper based substructure method on enhancing the identification accuracy.
In this paper, a loop substructure identification method is proposed to estimate the parameters of any story in a shear structure with measurements of only limited number of acceleration floors and unknown structural mass. A shear structure is divided into substructures consisting of a series of similar two-story standard substructures; two identification problems are formulated for the standard substructure using the cross power spectral densities (CPSD) of structural responses, each of which identifies the parameters of one story given that the parameters of the other are known. A loop identification scheme is proposed by connecting the two identification problems in a loop manner, forming a sequence of estimation problems to directly identify both story parameters of the standard substructure. If the structural masses are unknown, this loop identification method can still be applied to estimate mass normalized structural parameters as well as the relative mass distribution of the structure. The convergence condition is derived for the loop substructure identification, showing that the loop identification sequence is conditionally converged and some structural responses play a crucial role in determining the convergence. To achieve convergent identification results, a reference selection method is proposed, which uses a synthesized response, formed by a linear combination of the measured structural responses, as the reference response to calculate the CPSD and perform the loop substructure identification. A 20-story shear building is used to verify the convergence condition and to demonstrate that the proposed reference selection method does provide the converged and accurate estimation results.
In this paper, a new approach to identify the source location is proposed by exploiting the compressive sensing theory, which indicates that sparse or compressible signals can be recovered using just a few measurement. A square grid configuration plate with some piezoelectric actuator and sensor is used to verify the proposed approach. The grid is used to sweep across the plate to identify the location of source. Piezoelectric actuator placed on the plate is used to excite waves, and the signals of waves received at some sensors. The sensor locations are known, however, the source location need not be known. The candidate source locations are suitably chosen grid on the surface of plate. Sensing matrix which is related to the locations of source and sensor can be calculated at each sensor. Then, the proposed approach used the received signal strengths to locate the source by minimizing the ℓ1-norm of the sparse matrix in the discrete spatial domain based on the concept of compressive sensing (CS). The simulation results show the proposed method achieves a high level of localization accuracy.
Sparsity constraints are now very popular to regularize inverse problems in the field of applied mathematics. Structural damage identification is a typical inverse problem of structural dynamics and Structural damage is a spatial sparse phenomenon, i.e., structural damage occurs, only part of elements or substructures are damaged. In this paper, a structural damage identification method based on the substructure-based sensitivity analysis and the sparse constraints regularization is proposed. Substructure sensitivity analysis, the establishment of structural damage stiffness parameter variation and change of modal parameters of linear equations between the measured degrees of freedom is limited, the equations for a morbid equation. The introduction of structural damage sparsity conditions, to minimize the l1 norm optimization solution. The numerical example of the 20 bay-truss structure with considering measurement noise, incomplete of measurements and multi-damage cases are carried out. The effects of number sensor and layout to the identification results are also investigated. The results indicated that the damage locations and extents can be effectively identified by the proposed method. Additionally, the sensor location can be random arrangement, which has great significance to the sensor placement of the actual structural health monitoring because robust structural damage identification also can be obtained even a few of sensor are failure.
A moving loads distribution identification method for cable-stayed bridges based on compressive sampling (CS) technique is proposed. CS is a technique for obtaining sparse signal representations to underdetermined linear measurement equations. In this paper, CS is employed to localize moving loads of cable-stayed bridges by limit cable force measurements. First, a vehicle-bridge model for cable-stayed bridges is presented. Then the relationship between the cable force and moving loads is constructed based on the influence lines. With the hypothesis of sparsity distribution of vehicles on bridge deck (which is practical for long-span bridges), the moving loads are identified by minimizing the ‘l2-norm of the difference between the observed and simulated cable forces caused by moving vehicles penalized by the ‘l1-norm’ of the moving load vector. The resultant minimization problem is convex and can be solved efficiently. A numerical example of a real cable-stayed bridge is carried out to verify the proposed method. The robustness and accuracy of the identification approach with limit cable force measurement for multi-vehicle spatial localization are validated.
A data-driven approach for earthquake damage detection and localization in multi-degree of freedom
(MDOF) system subjected to strong ground motion is proposed. The new method is based on the
combination of wavelet analysis and fractal characteristics. The box counting method is employed to
obtain the fractal dimension of the time frequency distribution within the first natural frequency. It is
verified that the proposed fractal dimensions at each DOF of linear system are identical, while the
fractal dimension at the DOFs with nonlinearity will be different from those at the DOFs with linearity.
Therefore, the nonlinearity or weakness of the structure caused by strong ground motion can be
detected and localized through comparing the fractal dimensions at the measured DOFs. The numerical
simulation on a three-bay sixteen-story moment resist frame shows that the aforementioned approach is
capable of detecting and localizing seismic damage.
KEYWORDS: Sensors, Bridges, Sensor networks, Data acquisition, Structural health monitoring, Data communications, Signal attenuation, Wireless communications, Wavelets, Reliability
In a wireless sensor network, data loss often occurs during the data transmission between wireless sensor nodes and the
base station, which decreases the communication reliability in wireless sensor network applications. Errors caused by
data loss inevitably affect the data analysis of the structure and subsequent decision making. This paper proposed an
approach to recover lost data in a wireless sensor network based on the compressive sampling (CS) technique. The main
idea in this approach is to project the transmitted data from x onto y, where y is the linear projection of x on a random
matrix. The data vector y is permitted to lose part of the original data x in wireless transmissions between the sensor
nodes and the base station. After the base station receives the imperfect data, the original data vector x can be
reconstructed based on the data y using the CS method. The acceleration data collected from the vibration test of
Shandong Harbin Sifangtai Bridge by wireless sensors is used to analyze the data loss recovery ability of the proposed
method.
Optimal sensor placement is key issues of structures health monitoring (SHM). In study of sensor placement, the main
achievement focus on optimal criterions of sensor locations based on modal test, while optimal criterions of sensor
locations based on damage identification, optimal method of sensor locations and optimal sensor number should be
investigated further.
In this study, a novel optimal sensor placement strategy based on sensitivity is proposed. Optimal sensor placement
based on sensitivity analysis is an alternation method to consider damage identification. The basic idea of the proposed
methodology is that influence range of different damage parameters is different.
First, damage sensitivities in every element based on modal parameters are calculated. Then the elements that are
sensitive to damage are selected. According to the detection of damage sensitivity in these elements, minimums number
can be found by sensitivity. At last, the elements that are not selected are considered as not sensitive to modal parameters
and would be placed the strain sensors.
Numerical simulation of a three-dimensional truss structure is implemented to evaluate the minimum sensor number of
different damage parameters according to the above methods. Moreover, damage location can be detected under singledamage
situation and the element with most severe damage can be identified in multi-damage case using the proposed
sensor placement.
This paper presents a novel approach to detect structural damage combining non-negative matrix factorization (NMF)
and relevance vector machine (RVM). Firstly, the time history of acceleration signal are decomposed using the wavelet
packet transform to extract wavelet packet node energy as the damage feature, and construct a non-negative matrix using
the wavelet packet node energy index of all time history of acceleration data measured by multiple accelerometers
installed on the different locations of structure. Secondly, for increasing the damage detection accuracy, the dimension of
the feature non-negative matrix is reduced by NMF techniques and new representation of this matrix is obtained. Lastly,
RVM, a powerful tool for classification and regression, is used to detect the location of potential damage from the
reduced damage feature matrix. Numerical study on the Binzhou Yellow River Highway Bridge is carried out to illustrate the ability of the proposed method in damage detection.
Structural health monitoring (SHM) is regarded as an effective technique for structural damage diagnosis, safety
and integrity assessment and service life evaluation. SHM techniques based on vibration modal parameters are ineffective
for space structure health maintenance and the statistical process control (SPC) technique is a simple and
effective tool to monitor the operational process of structures. Therefore, employing strain measurements from optical
fiber Bragg grating (OFBG) sensors, the Johnson transformation based SPC is proposed to monitor structural health state
and some unexpected excitements on line in this paper. The large and complicated space structure-the China National
Aquatics Center is employed as an example to verify the proposed method in both numerical and experimental aspects. It
is found that the Johnson transformation can effectively improve the quality of SPC for SHM process, and it can clearly
and effectively monitor structural health state and detect the unexpected external load happened in structures.
KEYWORDS: Probability theory, Damage detection, Structural health monitoring, Finite element methods, Stochastic processes, Civil engineering, Error analysis, Statistical methods, Computer simulations, Monte Carlo methods
In this paper, a Dempster-Shafer evidence theory based approach for structural health monitoring is presented. Firstly,
Bayesian method is employed to calculate the damage probabilities of substructures using each data set measured from
the monitored structure, and the damage probabilities of substructures are transformed to damage basic probability
assessments which used in evidence theory. Then the Dempster-Shafer evidence theory is employed to combine the
individual damage basic probability assessments for getting the last damage detection results. With considering multi-sensors
data including acceleration and strain, and measurement noise the numerical studies on a 14-bay planar rigid
frame structure are carried out. The results indicate that the damage detection results obtained by combining the damage
basic probability assessments from each test data are improved compared with the individual results obtained just by
each test data separately.
Vibration-based damage identification is a useful tool for structural health monitoring. But, the damage detection results
always have uncertainty because of the measurement noise, modeling error and environment changes. In this paper,
information fusion based on D-S (Dempster-Shafer) evidence theory and Shannon entropy are employed for decreasing
the uncertainty and improving accuracy of damage identification. Regarding that the multiple evidence from different
information sources are different importance and not all the evidences are effective for the final decision. The different
importance of the evidences is considered by assigning weighting coefficient. Shannon entropy is a measurement of
uncertainty. In this paper it is employed to measure the uncertainty of damage identification results. The first step of the
procedure is training several artificial neural networks with different input parameters to obtain the damage decisions
respectively. Second, weighing coefficients are assigned to neural networks according to the reliability of the neural
networks. The Genetic Algorithm is employed to optimize the weighing coefficients. Third, the weighted decisions are
assigned to information fusion center. And in fusion center, a selective fusion method is proposed. Numerical studies on
the Binzhou Yellow River Highway Bridge are carried out. The results indicate that the method proposed can improve
the damage identification accuracy and increase the reliability of damage identification to compare with the method by
neural networks alone.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.