SENSOR BASED HEALTH MONITORING OF STRUCTURES By, M. Mayur, Siddharth Institute of Engineering and Technology, Puttur.Abstract:Structure is an element composing of manycomponents such as beams, columns, roofs,slabs, foundations and basements. Withoutbeams and columns, no structure is able tostand on ground. But these structures alsodamage due to temperature conditions theyexpose, mismanagement during constructionand lack of quality of control inconstruction. The damage is defined aschanges to the material or geometricproperties of a structural system, includingchanges to the boundary conditions andsystem connectivity, which adversely affectthe system’s performance. The SHM processinvolves the observation of a system overtime using periodically sampled dynamicresponse measurements from an array ofsensors, the extraction of damage-sensitivefeatures from these measurements, and thestatistical analysis of these features to Introduction:determine the current state of system health.After extreme events, such as earthquakes or The process of implementing a damageblast loading, SHM is used for rapid detection and characterization strategy forcondition screening and aims to provide, in engineering structures is referred to asnear real time, reliable information Structural Health Monitoring (SHM).regarding the integrity of the structure. Qualitative and non-continuous methods have long been used to evaluate structures for their capacity to serve their intended purpose. Since the beginning of the 19th century, railroad wheel-tappers have used the sound of a hammer striking the train wheel to evaluate if damage was present. In
rotating machinery, vibration monitoring has Operational Evaluationbeen used for decades as a performanceevaluation technique. In the last ten to Operational evaluation attempts to answerfifteen years, SHM technologies have four questions regarding the implementationemerged creating an exciting new field of a damage identification capability:within various branches of engineering.Academic conferences and scientific What are the life-safety and/orjournals have been established during this economic justification fortime that specifically focuses on SHM. performing the SHM?These technologies are currently becoming How is damage defined for theincreasingly common. system being investigated and, for multiple damage possibilities, which cases are of the most concern? What are the conditions, both operational and environmental, under which the system to be monitored functions? What are the limitations on acquiring data in the operational environment? Data Acquisition, Normalization andParadigm approach in SHM: CleansingThe paradigm approach of an SHM is The data acquisition portion of the SHMmainly divided in to four parts namely: process involves selecting the excitation methods, the sensor types, number and Operational Evaluation, locations, and the data Data Acquisition and Cleansing, acquisition/storage/transmittal hardware. Feature Extraction and Data Again, this process will be application Compression, and specific. Economic considerations will play Statistical Model Development for a major role in making these decisions. The Feature Discrimination. intervals at which data should be collected is another consideration that must be When one attempts to apply this addressed. paradigm to data from real world structures, it quickly becomes apparent Because data can be measured under varying that the ability to cleanse, compress, conditions, the ability to normalize the data normalize and fuse data to account for becomes very important to the damage operational and environmental identification process. As it applies to SHM, variability is a key implementation issue. data normalization is the process of These processes can be implemented separating changes in sensor reading caused through hardware or software and, in by damage from those caused by varying general, some combination of these two operational and environmental conditions. approaches will be used. One of the most common procedures is to
normalize the measured responses by the features identified from the undamaged andmeasured inputs. When environmental or damaged system. The use of analytical toolsoperational variability is an issue, the need such as experimentally-validated finitecan arise to normalize the data in some element models can be a great asset in thistemporal fashion to facilitate the comparison process. In many cases the analytical toolsof data measured at similar times of an are used to perform numerical experimentsenvironmental or operational cycle. Sources where the flaws are introduced throughof variability in the data acquisition process computer simulation. Damage accumulationand with the system being monitored need to testing, during which significant structuralbe identified and minimized to the extent components of the system under study arepossible. In general, not all sources of degraded by subjecting them to realisticvariability can be eliminated. Therefore, it is loading conditions, can also be used tonecessary to make the appropriate identify appropriate features. This processmeasurements such that these sources can be may involve induced-damage testing,statistically quantified. Variability can arise fatigue testing, corrosion growth, orfrom changing environmental and test temperature cycling to accumulate certainconditions, changes in the data reduction types of damage in an accelerated fashion.process, and unit-to-unit inconsistencies. Insight into the appropriate features can be gained from several types of analytical andFeature Extraction and Data experimental studies as described above andCompression is usually the result of information obtained from some combination of these studies.The area of the SHM process that receivesthe most attention in the technical literature Statistical Model Developmentis the identification of data features thatallows one to distinguish between the The portion of the SHM process that hasundamaged and damaged structure. Inherent received the least attention in the technicalin this feature selection process is the literature is the development of statisticalcondensation of the data. The best features models for discrimination between featuresfor damage identification are, again, from the undamaged and damagedapplication specific. structures. Statistical model development is concerned with the implementation of theOne of the most common feature extraction algorithms that operate on the extractedmethods is based on correlating measured features to quantify the damage state of thesystem response quantities, such a vibration structure. The algorithms used in statisticalamplitude or frequency, with the first-hand model development usually fall into threeobservations of the degrading system. categories. When data are available fromAnother method of developing features for both the undamaged and damaged structure,damage identification is to apply engineered the statistical pattern recognition algorithmsflaws, similar to ones expected in actual fall into the general classification referred tooperating conditions, to systems and develop as supervised learning. Group classificationan initial understanding of the parameters and regression analysis are categories ofthat are sensitive to the expected damage. supervised learning algorithms.The flawed system can also be used to Unsupervised learning refers to algorithmsvalidate that the diagnostic measurements that are applied to data not containingare sensitive enough to distinguish between examples from the damaged structure.
Outlier or novelty detection is the primary • Principle IV (a): Sensors cannotclass of algorithms applied in unsupervised measure damage. Feature extractionlearning applications. All of the algorithms through signal processing andanalyze statistical distributions of the statistical classification is necessarymeasured or derived features to enhance the to convert sensor data into damagedamage identification process. information; • Principle IV (b): Without intelligentIn total, feature extraction, the more sensitive a measurement is to damage, theOperation evaluation gives the conditions of more sensitive it is to changingSHM, operational and environmental conditions;Data Acquisition gives the number and types • Principle V: The length- and time-of sensors to be introduced in buildings, scales associated with damage initiation and evolutions dictate theFeature extraction gives the technical required properties of the SHMliterature to distinguish between damaged sensing system;and non damaged items of buildings, • Principle VI: There is a trade-off between the sensitivity to damage ofStatistical Model Development is used for an algorithm and its noise rejectiondetermining damaged and undamaged capability;structures. • Principle VII: The size of damage that can be detected from changes inPrinciples of SHM: system dynamics is inversely proportional to the frequency rangeBased on the extensive literature that has of excitation.developed on SHM over the last 20 years, itcan be argued that this field has matured to So far, we have known about SHM.the point where several fundamentalPrinciples, or general principles, have Let us know about it in a deepemerged. manner something about Components of SHM. • Principle I: All materials have inherent laws or defects; Components of SHM: • Principle II: The assessment of damage requires a comparison Structure between two system states; Sensors • Principle III: Identifying the Data acquisition systems existence and location of damage Data management can be done in an unsupervised Data transfer learning mode, but identifying the Data interpretation and diagnosis. type of damage present and the damage severity can generally only be done in a supervised learning mode; Data Interpretation and Diagnosis systems consist of:
1. System Identification, measured. Examples of this include 2. Structural model update, temperature, light intensity, gas pressure, 3. Structural condition assessment, fluid flow, and force. 4. Prediction of remaining service life. Data management:Sensors: Data management comprises all theSensors are a device that measures a disciplines related to managing data as aphysical quantity and converts it in to a valuable resource. The official definitionsignal that can be measured by an provided by DAMA International, theinstrument or by an observer. A sensor is a professional organization for those in thedevice which receives and responds to a data management profession, is: "Datasignal. A good sensor obeys the following Resource Management is the developmentrules: and execution of architectures, policies, practices and procedures that properly • Is sensitive to the measured property manage the full data lifecycle needs of an • Is insensitive to any other property enterprise." likely to be encountered in its application Data transfer systems are used to transfer the • Does not influence the measured data to systems which help in predicting the property. failures of structures.Data Acquisition Systems: StructureData acquisition is the process of sampling Conceptually, an accelerometer behaves as asignals that measure real world physical damped mass on a spring. When theconditions and converting the resulting accelerometer experiences acceleration, thesamples into digital numeric values that can mass is displaced to the point that the springbe manipulated by a computer. is able to accelerate the mass at the same rate as the casing. The displacement is thenThis includes: measured to give the acceleration. • Sensors that convert physical In commercial devices, piezoelectric, parameters to electrical signals. piezoresistive and capacitive components • Signal conditioning circuitry to are commonly used to convert the convert sensor signals into a form mechanical motion into an electrical signal. that can be converted to digital Piezoelectric accelerometers rely on values. piezoceramics (e.g. lead zirconate titanate) • Analog-to-digital converters, which or single crystals (e.g. quartz, tourmaline). convert conditioned sensor signals to They are unmatched in terms of their upper digital values. frequency range, low packaged weight and high temperature range. Piezoresistive accelerometers are preferred in high shock applications. Capacitive accelerometersData acquisition begins with the physical typically use a silicon micro-machinedphenomenon or physical property to be sensing element. Their performance is
superior in the low frequency range and they of the die. By integrating two devicescan be operated in servo mode to achieve perpendicularly on a single die a two-axishigh stability and linearity. accelerometer can be made. By adding an additional out-of-plane device three axes canModern accelerometers are often small be measured. Such a combination alwaysmicro electro-mechanical systems (MEMS), has a much lower misalignment error thanand are indeed the simplest MEMS devices three discrete models combined afterpossible, consisting of little more than a packaging.cantilever beam with a proof mass (alsoknown as seismic mass). Damping results Micromechanical accelerometers arefrom the residual gas sealed in the device. available in a wide variety of measuringAs long as the Q-factor is not too low, ranges, reaching up to thousands of gs. Thedamping does not result in a lower designer must make a compromise betweensensitivity. sensitivity and the maximum acceleration that can be measured.Under the influence of external accelerationsthe proof mass deflects from its neutral Building and structural monitoringposition. This deflection is measured in ananalog or digital manner. Most commonly, Accelerometers are used to measure thethe capacitance between a set of fixed beams motion and vibration of a structure that isand a set of beams attached to the proof exposed to dynamic loads. Dynamic loadsmass is measured. This method is simple, originate from a variety of sourcesreliable, and inexpensive. Integrating including:piezoresistors in the springs to detect springdeformation, and thus deflection, is a good • Human activities - walking, running,alternative, although a few more process dancing or skippingsteps are needed during the fabrication • Working machines - inside asequence. For very high sensitivities building or in the surrounding areaquantum tunneling is also used; this requires • Construction work - driving piles,a dedicated process making it very demolition, drilling and excavatingexpensive. Optical measurement has been • Moving loads on bridgesdemonstrated on laboratory scale. • Vehicle collisions • Impact loads - falling debrisAnother, far less common, type of MEMS- • Concussion loads - internal andbased accelerometer contains a small heater external explosionsat the bottom of a very small dome, which • Collapse of structural elementsheats the air inside the dome to cause it to • Wind loads and wind gustsrise. A thermocouple on the dome • Air blast pressuredetermines where the heated air reaches the • Loss of support because of grounddome and the deflection off the center is a failuremeasure of the acceleration applied to the • Earthquakes and aftershockssensor. Measuring and recording how a structureMost micromechanical accelerometers responds to these inputs is critical foroperate in-plane, that is, they are designed to assessing the safety and viability of abe sensitive only to a direction in the plane
structure. This type of monitoring is called information of the structural behavior ofDynamic Monitoring. bridges obtained from the monitoring system, maintenance costs could also beWIRELESS MONITORING reduced, since inspection methodsTECHNIQUES BASED ON MEMS (addressed i.e. in the following chapter) can be applied more efficiently. Only afterExisting monitoring systems use traditional certain changes in the structural behaviorwired sensor technologies and several other have been identified, an inspection (eitherdevices that are time consuming to install by means of non-destructive testing or visualand relatively expensive (compared to the methods) is necessary and proper repairvalue of the structure). They are using large could be done right after the occurrence ofnumber of sensors (i. e. more than ten) are the defect. This reduces the risk of furtherexpensive and will therefore be installed damage.only on a few bridges. A wirelessmonitoring system with MEMS (Micro- The analysis of measured data and theElectro-Mechanical-Systems) sensors could knowledge of continuous changes ofreduce these costs significantly. MEMS are structural behavior will also improve the lifesmall integrated devices or systems that time prognosis of civil structures reducingcombine electrical and mechanical the overall maintenance costs of buildingscomponents that could be produced for less and transport networks. Data has to bethan 50 € each. The principle of such a continuously transmitted (e.g. using thesystem is shown in the scheme given in Fig. internet) to the supervisor. Each sensor1. device (mote), which is itself a complete, small measurement and communication system, has to be power and cost optimized. Using multi-hop techniques, the data of the sensor network has to be transmitted over short distances of some 10 m to a base station on site. There the data items are collected and stored in a data base for subsequent analysis. This data can then be accessed by a remote user. If the central unit detects a hazardous condition by analyzing the data, it has to raise an alarm message. The central unit also allows for wirelessCurrently, a wireless sensor node with such administration, calibration anda MEMS sensor could be fabricated at a reprogramming of the sensor nodes in orderprice varying from 100 to about 400 € and to keep the whole system flexible. Eachfuture developments show the potential for mote is composed of one or more sensors, aprices of only a few Euro. Monitoring data acquisition and processing unit, asystems equipped with MEMS sensors and wireless transceiver and a battery as powerwireless communication can reduce the supply (Fig. 2, right) [3, 4]. The acquisitioncosts to a small percentage of a conventional and processing unit usually is equipped withmonitoring system and therefore will a low power microcontroller offering anincrease its application not only in integrated analogue to digital convertermonitoring bridges. Due to the detailed (ADC) and sufficient data memory (RAM)
to store the measurements. This unit alsoincorporates signal conditioning circuitryinterfacing the sensors to the ADC. In thefollowing sections, some components arementioned, but a more detailed descriptionis given elsewhere. An example of Micro machined Silicon sensor.A typical example of hybrid sensor systemfor wireless MEMS and DMS sensor data. An example showing monitoring of dams.A diagram showing sensors in structures. An example showing sensors in beams.
It is a typical example showing electricalgenerator and a sensor for health monitoring A type of forest based sensor for trees.of systems.An example of sensor based health An example of dam’s health in China.monitoring of structures.
A perfect Silicon Sensor for StructuralHealth Monitoring.Conclusion:The inspection of building structures andespecially of bridges is mainly done visuallynowadays. Therefore, the condition of thestructure is examined from the surface andthe interpretation and assessment is based onthe level of experience of the engineers. Anapproach to continuous structural healthmonitoring techniques based on wirelesssensor networks were presented, whichprovide data from the inside of a structure tobetter understand its structural performanceand to predict its durability and remaininglife time. Using this technique, monitoringof large structures in civil engineeringbecomes very efficient. . Essential is that thenew system provides a more reliable impactgeneration.