Patent Application of

James C. Elliott, Spencer B. Graves, and Sam Kovnat

TITLE: EQUIPMENT AND SYSTEM FOR STRUCTURE INSPECTION AND MONITORING

CROSS-REFERENCE TO RELATED APPLICATIONS: This is a PCT Application which claims priority to U.S. Provisional Application No. 61/231 ,460, filed August 5, 2009 and U.S. Provisional Application No. 61/236,900, filed August 26, 2009. These are incorporated herein by reference in their entireties.

FEDERALLY SPONSORED RESEARCH: Not applicable. SEQUENCE LISTING OR PROGRAM: Not applicable.

BACKGROUND— FIELD: This application relates to inspection and monitoring of structures such as bridges, towers, power lines, tunnels, steel storage devices such as oil tanks and water tanks, transportation containers, buildings, ceilings, roofs, walls, frames, stadiums (including seating), boats, ships, oil and gas platforms (both land- and sea-based), scaffolding, staging, dams, cliff and canyon walls, volcanoes, and other objects.

BACKGROUND - STATE OF THE ART:

This discussion of the state of the art is organized using the following categories:

• Introduction

• Current bridge inspection practices

• Remote access technologies

• Static, non-destructive testing • Sensors for real time data acquisition

• Monitoring technologies BACKGROUND - INTRODUCTION

Systems for inspection and monitoring can be classified as occasional and real time. Some occasional activities are scheduled long in advance; others occur in reaction to events that raise questions about the performance of a particular structure. Current bridge inspection practices, remote access technologies, and non-destructive testing generally involve occasional data collection. The discussions of monitoring and real time sensing falls into the latter "real time" category. There is, however, some overlap between these categories, and any real time activity will typically be used in conjunction with occasional reviews.

Scannell (2005) disclosed real time monitoring and Harrie et al. (2006) disclosed occasional monitoring. However, neither disclosure seems adequate to teach one skilled in the art how to solve certain practical problems in structural health monitoring for which certain valuable and unobvious advances are disclosed herein.

Scannell (2005) teaches methods for real time monitoring of structures on or over hydrological features. These are extremely important, because floods and scour account for roughly half of bridge failures in two recent studies (Wardhana and Hadipriono 2003; Biezma and Schanack 2007). Scannell claims, "A system to present data about at least one artificial structure in or over at least one hydrological feature, the system comprising" a database, a data source, and a user interface for presenting useful information about the level of potential risks to said hydrological feature, emphasizing the level of certain data elements relative to thresholds. This system focuses in particular on use of data on weather, geology and geography as well as certain existing technologies such as scour collars and sonic fathometers. This patent does not cover certain useful and unobvious innovations for (a) measuring deflections of a structure subject to varying load nor (b) data compression to reduce the expense of data communications and storage.

We found interesting the discussion of thresholds in the claims of Scannell (2005). Thresholds have been used for centuries in such a wide variety of ways and fields that the simple use of a threshold to trigger action seems to us rather obvious. For example, thresholds have been used in jurisprudence for centuries, e.g, in the distinction between petty and grand larceny or in deciding whether a person subject to inspection at an international border is carrying too much of something such as cash or alcohol. The concept of materiality on Accounting involves a threshold. Many organizations have "incoming inspection", comparing new materials against thresholds. Thresholds are involved in the valves releasing pressure from boilers on steam engines from at least the nineteenth century. For decades, such thresholds have been designed into the systems displaying a light on the dashboard of most automobiles when the oil pressure is too low or the engine temperature is too high. Thresholds have been used in terms of specification limits in engineering drawings for at least a century and probably much longer. Statistical control limits used in quality control since the 1920s are thresholds.

Thresholds were used in the design of dams before there were civil engineers. "The use of spillways probably dates back at least to the first great hydrological constructions of antiquity", according to the relevant French language Wikipedia entry (Wikipedia "deversoir").

There are many subtle and unobvious points on the selection of such thresholds, but Scannell (2005) made no contributions to this. Important contributions to threshold selection were made in the 1930s, 1940s and 1950s in the statistical literature on quality control (e.g., Wikipedia "statistical process control"), statistical hypothesis testing (e.g., Wikipedia "statistical hypothesis testing"), and decision theory (e.g., Wikipedia "decision theory"). These methods have been used routinely for decades in industrial engineering for quality control, in chemical engineering for managing the production in chemical plants, in electrical and electronic engineering to control defects and maximize yield in the production of electrical and electronic parts and assemblies, in agricultural engineering to develop better methods of horticulture and animal husbandry, in nuclear and mechanical engineering to manage production in power plants. There are 30-year-old cars on the road today with statistically designed thresholds in their on-board diagnostics (e.g., Box et al. 2000). A recent innovation regarding thresholds, dating from Benjamini and Hochberg ( 1995), is selecting thresholds to balance false discovery rate with false negatives.

Occasional inspections, as opposed to real time monitoring just discussed, was disclosed by Harrie et al. (2006) for a "Method and Arrangement for Inspection of an Object". This "method and arrangement" included an image recording unit, a display unit, and a positioning unit. The "object" could be virtually any large physical structure including a ship, a bridge, a building, and a marine or sea-based oil or gas platform.

The positioning unit could be a diver, van, helicopter, or a remotely operated vehicle. It could also include devices like ultrasonic transceivers mounted on a structure. The recording unit would typically include a database containing digitized drawings, photographs, sounds, and remarks entered via a user interface, all either either discrete or over time. However, accessing physical structures in air present unique challenges not discussed by this patent, for which useful and unobvious solutions are described in the present patent application.

Real time monitoring of the shape of a structure using ultrasonic transceivers seems not to have been considered, neither by Harrie et al. (2006) nor elsewhere in the literature. GPS has been used for such. Also, peer-to-peer networks using electromagnetic communications between nodes are used to identify the relative positions of different nodes (e.g., Twitchell 2010a, b, c). However, their use to monitor structures for changes in shape, i.e., deflection under changing load, have also not been discussed in the literature with which we are familiar. We have also not seen procedures for combining acoustic and electromagnetic monitoring to produce information superior to that available from either alone.

BACKGROUND - CURRENT BRIDGE INSPECTION PRACTICES:

There is ample documentation of massive deficiencies in current highway bridge inspection practices in the US. Current US law requires inspection every two years of all public highway bridges over 20 feet in length (NCHRP 2005, p. 5).

However, a review by MSNBC in Feb. 2008 found that only four states actually met that goal. One of these four was Tennessee, with over 19,000 bridges inspected by 17 inspection teams; thus, each team inspected on average over 2.2 bridges per work day (Zeyher 2008a, b).

Unfortunately, a failure to meet this desired inspection frequency is not the only problem: An important part of current procedures is visual inspection for cracks, which are difficult to see. A study of the reliability of human inspections found that 49 trained inspectors from across the US were able to detect only 3.9 percent of cracks in welded steel girders with fatigue-sensitive details; 96 percent of such cracks were missed (Washer 2003; Moore et al. 2001). Clearly, there are opportunities for new technology that can improve both the level of formal inspection coverage and the reliability of such inspections in detecting actual problems.

Even if human inspectors detected 100 percent of cracks in fatigue-sensitive details (rather than the 3.9 percent in the study just mentioed), that would not by itself tell us much about structural strength. Darwin et al. (2004) found that the number of cracks increased with compressive strength in monolithic and conventional overlay bridges, rather than decreased as one might expect. They also found that most cracks start early and grow with age. However, cracks do not grow linearly with age. Instead, they display jumps at apparently random points in time. In some cases, the jumps may be triggered by overload events of various kinds, e.g., overweight vehicles or high winds. In other cases, the appearance and growth of cracks may release internal stresses created by subtle, random variations in material strength and the curing process (cooling in steel, curing in concrete), possible generated as the structure redistributes the load in response to a new crack or crack extension.

Even if inspectors were capable of detecting and properly evaluating all cracks, that still would not solve the problem of fraudulent reports. In Feb. 2008 an inspection team supervisor in the Georgia Department of Transportation confessed to falsifying records after being asked how he could have possibly inspected up to 18 bridges on one day (Hart 2008). It is impossible to know whether and how many other fraudulent reports may have been filed without being caught.

Even if the inspections were perfectly performed, there remain the problem of the two-year time between inspections mandated for US highway bridges. On Sept. 7, 2009, the scheduled reopening of the San Francisco - Oakland Bay Bridge was delayed because an I-bar had fractured some time in the two years since the previous inspection; this fracture was caught because of other work done on that bridge (Wikipedia, "San Francisco - Oakland Bay Bridge). Fortunately, there was sufficient redundancy in the design that the fractured I-bar did not lead to a failure of the bridge. However, if that fractured I-bar had not been found when it was, the extra load on the others could have shortened their lives, leading to the fracture of a second, followed days rather than months later by the failure of a third, then an accelerated cascading failure dumping bridge and occupants into the San Francisco bay (Wikipedia, "Cascading Failure").

These inspections are required to follow procedures documented in the National Bridge Inspection Standards (NBIS; FHWA 2004). As suggested above, these standard are not uniformly applied, and when applied do not produce uniformly high quality results. "These evaluations typically consist of visual inspection and tap tests - listening to audible variations in response to tapping the bridge surface to determine if voids or de-bonding exist. ... Although based on current best practices, the inaccuracy of the current rating methods result in the retrofitting or replacement of many bridges that, in some cases, need not be retrofitted or replaced. Worse is the possibility that some bridges needing engineering renewal or replacement are not identified", according to Chang et al. (2003).

In sum, there are substantial opportunities for improving decisions regarding which bridges get funds for maintenance and replacement.

BACKGROUND - REMOTE ACCESS TECHNOLOGIES:

Relevant prior art for inspections includes both physical devices that make it easier for humans to get closer to relatively inaccessible features of current structures and measurement systems that could potentially outperform human inspectors in detecting problems. For example, Moog (1992) described a portable scaffolding to allow humans to more easily approach the underside of bridges. Many such systems go by names like "under bridge inspection unit". They are quite bulky, requiring a large truck for transportation. Moreover, their design makes them unsuitable for use with many bridges that do not have sufficient open space on the sides of the bridges to support their deployment.

Maciejczak (1988) proposed an unmanned assembly moving on a guided track carried by a space frame affixed to a structure. The assembly could carry cameras and other equipment for inspection, metrology and possibly minor repairs. These systems require expensive installation of the space frame with guided track.

DeVault et al. (1999) proposed a vehicle with six wheels for embracing and climbing a bridge piling. This will only climb a large cylindrical column and cannot move beyond any substantive obstruction. Alman (2006) proposed an unmanned propeller airplane requiring motion to generate lift to remain airborne. This aircraft can not remain airborne without continual forward motion at such a high speed that it would be difficult to get pictures as good as one could get with an aircraft capable of hovering. Moreover, other tasks requiring physical contact with the structure would be impossible with such a device.

Johnson (2007) proposed a ducted fan with counter rotating propellers, which however required a tether to provide power and communications. A device of this nature might be capable of close approach work for high resolution photographs or even physical contact with a structure. Similar devices have been prototyped for decades. Some are currently being marketed and developed by Trek Aerospace (2009). They may work fine in one position. However, they have problems traversing, because the ducts are not airfoils, and the competing demands of lift and thrust are difficult to reconcile. So far, such devices have failed to find a substantive market niche. This history raises questions about whether these devices will be able to perform sufficiently reliably to be considered viable alternatives to the devices discussed below.

Radio controlled helicopters might be used. Thunder Tiger (2009) markets a variety of such devices. Some are large enough to carry cameras or other metrology equipment with appropriate communications plus supplies and tools that could potentially be delivered to a human or some other more slowly moving device in a remote location. However, helicopters have wings (rotor blades) spinning at high speed and will be immediately destroyed if they touch any object such as a post, tree limb or the ground. Approaching too closely could pose an unacceptable risk of damage to the structure or to people or other property nearby. Without a system to substantially reduce the risk of such accidents, these helicopters would not be able to take high resolution photographs or perform other tasks requiring physical contact with the structure. Multiple publications have described automatic control systems for unmanned aerial vehicles (UAVs) specifically dedicated to bridge inspections, e.g., Metni and Hamel (2007) and Azinheira and Moutinho (2008). These systems could reduce the risk of an accident, thereby permitting slightly closer approaches. However, they are not currently capable of supporting the close approach and even physical contact proposed in this patent application.

Various kinds of climbing robots might be used. For example, Vex Robotics (2009) has a device using magnets in the tracks of a vehicle like a Caterpillar tractor or army tank. The magnets allow the vehicle to climb vertically on ferromagnetic substrates. However, it has so far not been able to transition from climbing vertically to hanging from the underside of a ferromagnetic structure. In addition, it requires a moderately smooth surface.

Other devices use "sticky foot" technology (Graham-Rowe 2007), using the same physical principles that allow insects to hang upside down from ceilings.

Presently available devices using this technology are able to climb on fairly clean surfaces. However, they would be close to useless with structures such as bridges where flaking paint, rust or other contaminants would quickly foul "sticky foot", destroying its stickiness.

The "RiSE" robot project (2009) uses gripper feet to climb trees and rough surfaces such as concrete. However, steel bridges might be too smooth for their grippers. Boston Dynamics (2010) displays a variety of possibilities on their web site, but much of that looked like virtual reality. A patent search for "Boston Dynamics" exposed both applications and issued patents, but not for technology that seemed relevant to the present application.

BACKGROUND - STATIC, NONDESTRUCTIVE TESTING:

Van der Auweraer and Peeters (2003) noted that, "Most currently used damage detection methods are visual or localised methods using acoustic, ultrasonic, magnetic field, X-ray or thermal principles. All these techniques require a priori knowledge of the vicinity of the damage." Medwick and Kaufman (2007) provided a recent discussion of new inspection technologies including "magnetorestrictive testing", "D- meter" ultrasonic testing, fiber optic borescopes, digital moisture meters (to help identify rotting timber), and innovations in design for inspectability. Rens et al. (2005) compared acoustic emission, electrical, magnetic, sonic, and surface hardness methods, impact-echo, radar, radiography, thermography, acoustic tomography, and ultrasonics. They provided an ultrasonic example produced using piezo-electric ultrasonic transducers. This example included 3-dimensional images of the internal integrity of concrete, crudely similar to the images produced by magnetic resonance imaging (MRl) and computed tomography (CT). The resolution of images produced by this technology would generally increase with increases in the number of transducers.

Methods using sound have been described by Costley et al. (2003), Huang (2001), Kennedy et al. (2004), and Rhazi (2006), among others. Parzichowski et al. (1007) described a "portable laser based ultrasonic" system.

Other methods use various electromagnetic principles including electrochemical fatigue sensors (Moshier and Berks 2009; Matech 2009), radar (Alongi and Alongi 1987), infrared (Bodkin 2000), X-rays (D'Ambrosio 2002), lasers (McGugin, et al. 2001 ) and magnetostriction (e.g., Livingston et al. 2002). There has been substantial development of ground penetrating radar for isolating defects in bridges. One example is a hand-held radar unit to detect problems in concrete bridges and piers (Brehm 2007). Tung et al. (2002) described a "mobile manipulator imaging system for bridge crack inspection", including a new pattern matching algorithm for processing binocular pictures from charge coupled device cameras.

For historic wrought iron bridges, Gordon and Knopf (2005) recommended metallurgical evaluations using, e.g., a microscope to identify iron samples that are too brittle or not sufficiently ductile.

Most of these technologies must either be incorporated into the structure when it is built or added to the structure later or carried repeatedly to the structure when measurement is desired. Some are sufficiently expensive that they will only be added locally for a short period of time.

BACKGROUND - SENSORS FOR REAL TIME DATA ACQUISITION:

An excellent overview of the state of the art for real time monitoring, especially of bridges, is provided by Aktan et al. (2003). Much of what they say could be applied to natural structures and constructed facilities other than bridges. Aktan et al. (2003, sec. 4.2) divides sensors into those measuring direct strain, linear displacement and position, temperature, acceleration, tilt, weight in motion, global position (GPS), acoustic emissions, and environmental conditions such as wind, humidity, and ambient temperature.

Direct strain measurement can be done using electrical resistance, vibrating wire, or structural moment detectors (e.g., Scott and Rhoades 1984), and various fiber optical systems such as fiber Bragg gratings and extrinsic Fabry-Perot interferometry. Svaty (1995, 1996) patented the use of strain gauges to evaluate resonant frequencies and modes, monitoring them over time for changes.

"The most commonly used and practical sensors for vibration testing are accelerometers" (Akten et al. 2003, sec. 4.2.4), though other techniques for vibration analysis have been proposed. For example, McGugin et al. (2001 ) proposed a laser system for this purpose.

Unfortunately, the interpretation of direct strain and vibration data is always questionable because of its essential reliance on strong assumptions that the existing condition is adequately modeled in the analysis. For example, such an analysis might NOT have detected the primary cause of the infamous collapse of the footbridge in the Hyatt Regency hotel in Kansas City, July 17, 1981 , killing 1 14 people and injuring over 200 others. The problem in that case was a critical deviation between design and construction (Wikipedia, "Hyatt Regency walkway collapse"). Strain and vibration analyses also may have difficulty adequately characterizing problems due to inhomogeneity of strength of materials whether due to irregularities in production or in the aging process.

Many of the problems with strain measurements and accelerometry can be overcome with linear displacement and position sensors, provided they actually measure distances that record how changes in loading change the shape of the structure independent of design details like the one that contributed to the Hyatt Regency bridge collapse previously mentioned. Linear displacement and position sensors include cable transducers (e.g., Brewer et al. 1961 ), LVDTs (e.g, Wikipedia "linear variable differential transformer"), vibrating wire crackmeters, and fiber optic displacement sensors of various kinds (e.g., Hodges 2004, 2003, 2002a, 2002b, 2001 , 2000; Metje et al. 2008; Pozzi et al. 2009; Glisic and Inaudi 2007). Other systems use lasers to detect motion (e.g., Savino 1989; Canty and Canty 1995) but may not have adequate resolution for many purposes.

Installed to measure critical distances, linear displacement / position sensors can potentially provide direct measures of the performance of the structure to variations in load largely independent of any assumptions about details of design or construction. Monitored over time, they can provide early warning of failure by noting movements in reaction to load that do not return to the previous unloaded position when the load is removed. Such measurements indicate that the load has exceeded the elastic limit of the structure, thereby causing permanent damage and reducing the ability of the structure to perform in the future. Linear displacement and position measurements are arguably the most important data that could be collected on a structure precisely because it is minimally impacted by any assumptions about design and construction. Linear displacement and position sensors might also provide information about problems with piers. For this, they may not be as sensitive as tiltmeters but might still provide substantial early warning of problems with scour or collisions. This has, however, rarely been attempted in the past, especially for bridges, because the obvious distance, e.g., from the midpoint of a span to the ground below (or any similar distance), cannot be measured directly: A bridge spans an obstruction that makes such measurements infeasible, except in rare cases.

Temperature is often measured to allow for corrections to be made in readings by temperature sensitive sensors and to estimate thermal expansion. Measuring other ambient weather conditions such as wind speed can be used to evaluate other loading effects on structures.

Acoustic emissions monitoring listens for the sounds made by, e.g., strands of cable breaking under load. Despite encouraging laboratory results and attempts to monitor acoustic emissions since the 1930s, acoustic emissions monitoring has yet to establish itself as a viable method for monitoring structural health (e.g., Rens et al. 2005).

Weigh in motion sensors are sometimes used to monitor the load on a structure in real time.

Tilt of piers can provide sensitive indications of problems with floods, scour, and collisions with ships, trains and trucks. Floods and scour were the primary cause for over half of the roughly 500 bridge failures in the US between 1989 and 2000 studied by Wardhana and Hadipriono (2003), and collisions accounted for another 12 percent (Wardhana and Hadipriono 2003, Table 5). Tiltmeters monitored in real time could provide almost instantaneous reporting of problems created by scour and collisions. In the most extreme cases of large ships removing a pier, there may not be time to react before the bridge collapses. More commonly, however, the scour or collision will first register as a relatively minor change in the tilt of the pier, which might grow initially at a sufficiently slow rate to allow the bridge to be closed and studied for structural damage. Without such real time monitoring, the bridge could proceed to failure before anyone becomes aware of the problem.

Nassif et al. (2005) found that laser Doppler vibrometry was as good as contact sensors, both linear variable differential transducers (LVDTs) and geophones, for monitoring bridge deflection and vibration. However, the current costs of laser Doppler vibrometry and related technologies may make them infeasible for many applications.

Global position sensors (GPS) have been used to monitor the position in space of particular points on a structure. Unfortunately, GPS systems of sufficient accuracy may be too expensive for routine use in structural health monitoring. In principle, however, they could provide some of the same measurements as more traditional linear displacement and position sensors.

Lopez at al. (2010) described the use of optical scanners, not necessarily using lasers, similar to an automated theodolite. Tobalske et al. (2007) tracked the flight of a hummingbird in 6 dimensions (3 for linear position plus yaw, pitch and roll) using multiple video cameras. In principle, either of these technologies could be used for physical infrastructure. However, it's not clear if they could produce the kinds of resolution needed without clear specification of reference points to track. Tobalske et al. (2007) had humans manually designate landmarks or reference point in a few frames from each camera. The results were acceptable for their purposes but may not be adequate for structural health monitoring. There is a substantial literature on photogrammetry (e.g., Wikipedia, "Photogrammetry") but little if any use in structural health monitoring; we believe this is because of the difficulties of obtaining sufficiently precise measurements without explicitly designated landmarks.

Almost a century ago, the problem of studying motion from photography was solved in pioneering contributions to the then-new field of industrial engineering by Lillian Gilbreth: She placed corner reflectors on various parts of her body and used a strobe light with slow film and a small aperture to trace her motions while moving around her kitchen in meal preparation. She used this information to redesign her kitchen. The use of comer reflectors for localization in digital video photogrammetry might provide an order of magnitude improvement in precision over what is achievable with other landmark registration techniques. However, we have not seen more recent documents using this technique.

BACKGROUND -- MONITORING TECHNOLOGIES:

All measurement includes measurement errors, and no measurement will be perfectly correlated with the probability of a subsequent failure of a structure monitored. For example, the level of deflections and vibration of a bridge depend on multiple factors such as traffic, wind and weather conditions in addition to details of design and construction. Many methods have been developed for monitoring processes generally, ranging from statistical control charts (e.g., Shewhart 1931 ) to exponentially weighted moving averages, more general Kalman filters and state space techniques to a variety of algorithms for artificial intelligence. Kalman filtering techniques have been extended to handle nonlinear state space approaches for assessing structural damage by Overbey and Todd (2008) and to particle filtering by Xue et al. (2009).

Designing a structural health monitoring system is similar to the problem of designing on-board diagnostics to detect malfunctions in the emission control systems of automobiles, described, e.g., by Box et al. (2000). Good monitor design requires clear definitions of both good and bad systems with an adequate understanding of the probability distribution of the data collected from both good and bad systems. This foundation can then be used to compare alternative monitors and even design monitors that are optimal under the given specifications of good and bad.

A major design consideration for any system to monitor a large structure is the cost of data communications and storage. Existing computer and sensor technology support collecting data much faster than is needed most of the time and faster than can be justified economically generally. The key question in the selection of sampling rate, data transfer rate and data storage is the information contained in the data collected. There is a substantial literature on adaptive data compression, but we have so far seen nothing that combines physical considerations with statistical data analysis. This, we believe, is key to designing optimally efficient data compression and storage algorithms.

For example, modern metrology can in many cases provide more digits than the available instruments can reliably measure. Standard lossless data compression algorithms (Wikipedia "data compression") will in many cases fail to achieve much compression, because they faithfully store the noise with the information desired. Lossy algorithms can do much better, provided they have a clear understanding of the underlying dynamics of the process being measured and the nature of the noise in the measurements. Much of the work in lossy data compression seems to have focused on compressing vido or audio so humans cannot detect the loss (e.g., Wikipedia "lossy compression"). Recent work has described data compression using piecewise constant approximation (Lazaridis and Mehrotra 2003), Kolmogorov-Sanai entropy (Titchener 2008), a Markov expert system (Cheng and Mitsenmacher 2005), statistical moments (Choi and Sweetman 2009), nonparametric procedures (e.g., Ryabko 2009, 2008), autoregressive moving average summaries (Sridhar et al. 2009), Fourier analysis (e.g., Reddy et al. 2009), extrema (e.g., Fink and Gandhi 2007), neural networks (e.g., Izumi and Iiguni 2006), and time series data mining (Li 2010). However, we've seen no discussion of how use in designing a data compression algorithm (a) the dynamics of the physical system monitored and (b) the noise of the sensors with (c) the accuracy demands of the application.

In principle, piecewise constant approximations could be used in this way for data from accelerometers. However, Lazaridis and Mehrotra (2003), who discussed piecewise constant approximations, used using L _x error bounds; the use of L _∞ is equivalent to assuming double exponential noise, which is produced by no physical process we know.

Changes in temperature and displacement can often be modeled with second order differential equations plus measurement noise that may be a mixture of normal distributions but not a double exponential. Such second order dynamics could easily be modeled with as a hidden Markov process with a two- or three-dimensional state vector consisting of the position, velocity and possibly acceleration. A "Markov expert system" may include an such a model as an option but will in general waste resources, including communications bandwidth and data storage capacity, considering alternatives that may be physically impossible for the particular application. Information theory has shown itself to be extremely useful for data compression and communications, but we have so far seen no literature that appropriate considers the known physics of the structure and sensors in so-called "information" or entropy-based data compression and communications. Neural networks and "expert systems" may outperform a Kalman filter, for example that is a poor match to the physics. However, we would not expect artificial intelligence to perform as well as an algorithm that appropriately considers the physics of the application.

Systems for distributed Kalman filtering (e.g., Olfati-Saber 2007) can support models closer to the physics than the data compression algorithms we've seen.

However, such systems can be extremely difficult to design and use, because we must either specify the model entirely when we install the sensors or allow the system to be reprogrammed remotely. If the model is completely specified in advance, it may be not be feasible to modify the mathematics later to exploit improvements in our understanding of the behavior of the structure. If the system can be reprogrammed remotely, it increases the cost of the smart sensors and computers located with the structure and increases the risks of hacker attacks.

Decentralized systems for damage detection and isolation using vibration analysis have been described by Svaty (1995, 1996), Nagayama et al. (2009), and others.

A sufficiently detailed and accurate model of the actual physical system is useful no matter what kind of data are collected. Proper interpretation of data in deflection requires only a relatively crude model. Much more sophisticated models are virtually essential for proper interpretation of other kinds of data. Moustafa et al. (2010) advocated building such models using bond graphs.

Techniques for more automated structural / systems identification from data have been developed partly to respond to this need (e.g., GuI and Catbas 2008;

Saadat et al. 2007). Catbas et al. (2007) noted, however, that this is extremely difficult, especially for large structures. However, they used data on local strain, acceleration, wind and temperature. We believe that better results might often be obtainable using direct measures of physical deflection at key locations in place of indirect measures of strain and acceleration.

More generally, Carden and Fanning (2004, p. 18) observed that no algorithm has yet been proposed, which can be applied universally to identify any type of damage in any type of structure. This is especially true of the very popular vibration based analyses. For example, Humar et al. (2006, p. 215) observed that, "in practice, a number of difficulties persist in vibration-based damage identification. As a result, most of the damage identification algorithms fail when applied to practical civil engineering structures." Similarly, Farhey (2006) summarized 4 years of experience with a real time monitoring system installed on a new bridge. He noted that, "While the use of structural instrumentation and monitoring in other industries has been extensive, it is still considered logistically complicated, labor-intensive, time- consuming, and expensive for civil infrastructure applications. Although diagnostic evaluation offers potential advantages over conventional visual inspection, currently, this practice is not being used yet for routine management decisions. The use of structural diagnostic techniques is increasing, but only for important decision-making purposes, such as on high profile or critical structures". One contributor to this problem is the fact that temperature variations (daily and annual thermal cycling) can have a larger impact on structural dynamics (both resonant frequencies and mode shapes) than damage (Xu and Wu 2007; Chang et al. 2003).

The state of the art in this regard is illustrated by a private comment in 2009 from an engineering manager in the department of transportation of a major state: One bridge in his inventory had been instrumented with 150 accelerometers that produced zero useful information.

Overall, general theories often do not provide sufficient guidance for many specific applications, and the special cases that have been documented are not sufficiently general to be useful in many particular cases.

The above methods deal with less than half of bridge failures. Recent research reports have identified problems with hydrology as the primary cause of over half of bridge failures (Wardhana and Hadipriono 2003; Biezma and Schanack 2007). Blanchard (2003) identified "Field measurements of scour, especially abutment and contraction scour", "Correlation of easily observable variables and scour potential", "Time development of scour" and "Quantitative predictions of geomorphic change" as four often "research needs" identified by the TRB (1996); see also Lagasse, et al. (2006). Yu and Yu (2009) list "yardstick, ground penetrating radar, ultrasonic method, and fisher bulb" as existing alternatives for measuring scour, while describing in some detail the use of time domain reflectometry for monitoring scour.

WeatherData, Inc. holds a patent on a "method and apparatus for activating weather warning devices" (Smith 2002), and have filed applications on a "method and apparatus for activating warning devices" (Smith 2004) and "normalized and animated inundation maps" (Smith and Long 2004). These have been used to warn railroads and others of impending floods. A similar patent targeted for structures in or over hydrological features is Scannell (2005), mentioned above.

Radio frequency methods for localization and monitoring are disclosed in a series of patents, including Twitchell (2008, 2009, 2010a, b, c) and references cited therein. Methods for radio frequency monitoring of container contents are described by Twitchell (2008) with additional details for hazmat containers in Twitchell (2009). Refinements for radio frequency localization appear in Twitchell (2010a, b, c). The biggest gap in these methods is their failure to consider how to detect changes in the physical shape of a structure. Twitchell (2009) describes the use of data from accelerometers and a variety of other sensors such as biological, radiological or chemical sensors to monitor the condition of a mobile hazmat container. Twitchell (2010a, b, c) describe the use of radio frequency ranging extending GPS technology to GPS-denied areas but fails to note that this technology could be used to monitor for changes in the shape of a physical structure. In some cases, monitoring for changes in shape could be more informative or at least a useful confirmation of accelerometry results. Secondarily, a very important subset of what is described in these Twitchell patents and related prior art could also be achieved using acoustic methods or some combination of acoustics and radio frequency methods. Acoustics could provide clarity in areas where electromagetics may not be adequate, but we have found no discussion of these techniques in the prior art. Algorithms for translating pairwise distance measurements into multidimensional maps have been known since the 1950s with algorithms for multidimensional scaling, especially "metric multidimensional scaling" (Wikipedia, "Multidimensional Scaling").

BACKGROUND - ADVANTAGES:

Equipment and methods in this patent application make it easier and cheaper to evaluate the health of a physical structure. The first two embodiments described here support real time monitoring of the dynamic reactions of a structure to changes in load. If the structure does not return to its unloaded state after a load is applied, the owners of the structure can be notified immediately that the structure has either suffered permanent damage or is displaying hysteresis. Engineers familiar with the structure can determine how large a change could reasonably be attributed to hysteresis, with larger changes suggesting a need for a serious assessment of the remaining strength of the structure. They employ novel reflective means for landmarking or provide stable reference points not previously discussed in the prior art for measuring structural deflections under changing load.

Other embodiments provide easier access to parts of structures that are otherwise difficult to access. This makes it easier to take higher quality photographs than might otherwise be feasible for a comparable budget. It also makes it easier to install remote measurement equipment for monitoring the condition of the structures, which in turn will facilitate better decisions for maintenance and repair of such structures. These devices seem new and unobvious from our review of the prior art.

Another embodiment provides mean for compressing data in computers on or near the structure monitored to minimize the cost of data communications and storage, adjusting the data transmission and storage requirements to maintain an acceptable error level in ability to predict future observations: New data transmission and storage are triggered only when the previously available information are not adequate to predict the latest observation(s). The novelty here rests on using knowledge of the physics with the known strengths and limitations of the sensors to design the data compression algorithms.

BRIEF SUMMARY: We describe four machines and a method for data compression to reduce the costs of data communications and storage.

BRIEF SUMMARY: MACHINE l : DYNAMIC DEFLECTION METROLOGY, FIGS I A, I B, 1C

A machine to measure the deflection of a beam such as a bridge span is disclosed. The machine consists of (a) one or more linear displacement or position sensors measuring the distance between (b) an identified informative point on the structure and (c) a stable reference point. The said stable reference point is created by equipment connected to points on or near the structure that do not move under normal circumstances and hence provide the stability of the said stable reference point. If the said connection points on or near the structure actually move, such motion either suggests an event requiring an investigation by a human or are so rare that data indicating excessive deflection would still justify an investigation by a human.

BRIEF SUMMARY: MACHINE 2: DYNAMIC REMOTE OBSERVATION USING REFLECTIVE MEANS, FIGS 2A, 2B, 2C, 2D

A system to measure the movement of specified informative points on a structure such as a bridge, dam, hazmat facility, shipping container, marine oil or gas platform, cliff face or building is described. The systems consists of two parts: (a) reflective means attached to one or more informative points on the structure and (b) one or more reference devices determining locations by measuring some combination of range, azimuth and elevation between a sufficient number of pairs of reflective means and reference devices. The said reflective means could be a corner reflector or a transponder. The said reference device(s) could be a camera, optical scanner, laser or a transponder, similar to GPS but not necessarily using a satellite. The novelty here is in using these data to provide early warnings of changes in shape of the structure.

BRIEF SUMMARY: MACHINE 3: PROTECTED VTOL ₅ FIGS SA ₅ SB ₅ SC SD

One machine described herein is a VTOL (Vertical Take off and Landing) device, specifically a remotely controlled helicopter with protective bumper bar(s) to allow the device to be safely flown much closer to a physical structure than would be feasible without an unacceptable risk of damage to the structure, helicopter and anything nearby that might be damaged if the helicopter rotor struck part of the structure.

The VTOL (Vertical Take off and Landing) system described below is intended for use with radio controlled (RC) helicopters such as the Raptor 60 or 90 series marketed by Thunder Tiger (2009), though smaller or larger helicopters could be used depending on the weight of the specific payload (e.g., camera, communications equipment, other metrology equipment or installation tools) required to be carried to the desired remote location. [The 60 or 90 series belong to the .60 or .90 class, which refers to the size of helicopter that could be powered by (a) a two-cycle engine with .60 or .90 cubic inches displacement or (b) an electric motor of comparable power.]

Helicopters have wings (rotor blades) spinning at high speed and will be immediately destroyed if they touch any object such as a post, tree limb or the ground. An appropriate protection system will allow an RC helicopter to safely approach structures and objects while carrying other equipment, supplies or special systems under the control of an operator on the ground. The protection system is especially beneficial where the RC helicopter is carrying a camera or other technologies to examine structure integrity, and the operator may be concentrating on the structure condition and not notice an obstacle that would interfere with the moving rotor.

Generally, the greater the power available for the actual weight, the more applicable the use of this protection system.

A very small class RC helicopter might employ the minimal application of a few protectors such as an arrangement of three evenly spaced protection rods in order to provide crash protection during training operations. Eight or more protection rods could provide additional protection for larger RC helicopters such as those in the .90 class.

The novelty here is the bumper bars allowing much closer approach to a physical structure than would be possible using an unprotected helicopter.

BRIEF SUMMARY: MACHINE 4: CLIMBING ROBOT, FIGS 4A, 4B, 4C

Climbing robots can move from point to point through several different modes. One basic motion is similar to that of a two legged animal or person. On a surface that is horizontal or has only a modest grade, the motion depends on gravity and balance. While one leg is lifted, the other provides a fixed position, and the body adjusts to maintain balance. The lifted leg is then firmly placed on the ground or other surface, and the other leg is lifted while the balance is shifted to the leg at rest on the ground, again relying upon gravity. Repetition provides a continuation and where the placement of each leg (step) occurs it provides direction. A four legged animal will place two legs in position while moving two legs and the end result is the same. A robotic device can imitate the movement of a two, four, six or more legged animal or it can use wheels, rollers, or rolling track to accomplish the same movement, possibly using the physics of gravity.

Some specially designed robots can overcome gravity by moving about in an additional dimension such as changing from horizontal to vertical by using a vehicle with magnetic track (Vex Robotics 2009) or a technology called "sticky foot" (Graham-Rowe 2007). The magnetic track is normally structured like the track of a tank, and the direction of motion is changed by slowing or speeding up one side compared to the other. Sticky foot uses numerous suction cups or a very specialized material coated with a glue-like substance and can be structured like a track or a pad or as numerous pads. However structured, they are designed to allow temporary attachment while the robot moves in a horizontal or vertical direction or even suspended from the underside of the surface upon which it is moving. Magnetic track can be very effective. However, there must be a sufficiently strong attachment to allow the device to continue hanging. For example, an "I" beam presents a sufficient amount of flat surface underneath the beam but a very small amount of surface on the edge of the lower horizontal plate. In some cases, the geometry of the tracks and the "I" beam may not support a sufficiently strong attachment between the two to support the weight of the climbing robot. Similarly, a magnetic track would not be able to move along a surface that was very thin or move along a wire.

A robot with sticky feet could effectively be limited to moving on clean surfaces. If a robot with sticky foot were to move along a steel structure that contained rust, dirt, flaking paint or other contaminants it could quickly loose its stickiness due to contaminants adhering to the sticky surface, thereby reducing the strength of the bond with the structure below the minimum required to hold the device to the structure.

The proposed robot would have a structure and operation similar to the "RiSE" robot project (2009) which climbs vertical structures using multiple legs and specialized feet. The electromagnetic robot (EMR) disclosed here replaces at least some of the gripper feet with magnetic equipment. The magnets will allow more secure movement on a ferromagnetic structure, moving horizontally or vertically, right side up or up side down. Unlike the RiSE robot, the EMR will be able to park for an undetermined period of time of time consuming zero power to remain in place; the RiSE robot can park for only a limited period, because it requires power to maintain its grip on the surface upon which it is climbing. The magnetic feet will allow the EMR to work using substantially less power than for locomotion. It could collect data using metrology equipment carried to the remote location and relay it to someplace else. Or it could install other equipment to leave on the structure permanently. The power required for these metrology or installation tasks would generally be an order of magnitude less than the power required to climb.

The novelty here is the electromagnetic feet allowing the robot to climb the possibly uneven surface of structures with sufficient ferromagnetic content.

SUMMARY: METHOD 1 : DATA COLLECTION AND ANALYSIS SYSTEM, FIGS 5A, 5B, 5C

We also propose to use data collected either irregularly using metrology carried by devices such as the protected VTOL or the EMR or metrology units installed in relatively inaccessible locations to improve the capabilities to more accurately predict the probability of failure under normal and stress conditions such as floods, storm surges, high winds, and earthquakes. This information could further be used by appropriate decision makers to decide which structures should be repaired or replaced and which should received further evaluation before final decisions are made regarding which structures receive the limited funds available. These kinds of decisions are routinely made today using whatever data are available, mostly visual inspections. The availability of better, more informative data, possibly collected continuously, should provide a material improvement in the ability of managers to better respond to their mandates, including better data collection and maintenance of infrastructure with the funds available ~ and including an improved ability to defend their budgets to higher officials and politicians based on more vivid and convincing descriptions of the risks of working with smaller budgets. BRIEF SUMMARY: METHOD 2: TIME SERIES DATA COMPRESSION ₅ FIG O

We disclose herein a novel method of data compression that exploits the physics of the system monitored and the noise level of the sensors. Previous data

compression techniques exploit repetitive elements in the data that do use knowledge of physics and metrology noise.

Prior art in time series analysis includes filtering method for converting a number N of data points into an equal number of observations on an estimated state space representation plus error. The method disclosed here only saves the current state if the error in predicting the current observation from the last saved state exceeds some threshold.

This provides three important advantages over the ARMA(I , 1 ) procedure described by Sridhar et al. (2009). First, it includes a measure of the prediction error; Sridhar et al. (2009) simply assume that the system is always fine. Second, it adapts to the behavior of the structure, transmitting and storing more data when that is required to improve the predictions. Third, it supports the use of more general statistical models, potentially more tied to the physics of the structure that a purely statistical ARMA(I , 1 ) model. Beyond these, it may also optionally support queuing data transmissions for off-peak times of day or week when the costs of data communications may be less.

DRAWINGS-Figures

Figs 1 A-I C describe the dynamic deflection measurement machine. Figs 2A- 2D describes various illustrative applications of our more general machine for tracking changes in the locations of one or more identified informative points on a structure.

Figs 3A-3D show various aspects of the VTOL device, emphasizing the novel protective bumper bar(s). The device will also include a high resolution camera or some other payload (not shown) for collecting data on the condition of an existing structure of interest at a lower cost than known alternative technologies (unless the structure included such metrology equipment in its design and construction).

Figs 4A-4D show various aspects of magnetic feet used with a climbing robot.

Figs 5A-5C show a method of using the VTOL device with either an EMR or with a similar climbing robot or with a human to either collect more data than might be cost effective with only one of these approaches or to support cost effective installation of data collection equipment designed to remain with the structure indefinitely.

Fig 6 summarizes our data compression method. DRAWINGS-Reference Numerals:

1 ) Protective Rod (Bumper Bar)

2) Rotor Blades

3) Tail Rotor

4) Fuselage

5) Engine

21 ) Magnetic Foot

22) Cap

23) Positioning Spring

24) Electromagnet shell

25) Permanent Magnet 26) Non-Magnetic Foot Pad/Cushion

27) Robot Leg/Arm

28) Robot

29) Recessed Neck Trap 10I) A span of a bridge

102) bridge deck

103) identified informative point, typically in the middle longitudinally of a span

104) stable reference point

105) nominally stable points on the structure or the environment

106) connections of fixed length to maintain constant the distance between 104 and 105

107) piers or abutments of a bridge

1 1 1) metrology mounting bracket affixed to the underside of the bridge deck holding linear displacement / position sensor(s) in fixed position(s) relative to the the identified informative point, 103.

121) inner tube

122) outer tube

123) mount for a linear displacement / position sensor measuring vertical displacement of the identified informative point 103 relative to the stable reference point 104. 124) connection means between the linear displacement / position sensor mounted at 123 and the outer tube 122.

125) mount for an optional linear displacement / position sensor measuring horizontal displacement of the inner (121 ) and outer (122) tube assembly, indicating a change in one of the piers or abutments 107 relative to the other.

126) connection means between the linear displacement / position sensor mounted at 125 and the outer tube 122.

201) a dam

202) identified informative points on the physical structure (a dam in Fig 2A, bridge in Fig 2B, etc.)

203) monitoring devices for tracking changes in the locations of identified informative points 202.

21 1) bridge

221) marine oil or gas drilling or production platform

222) identified informative points on a physical structure equipped with a transponder for peer-to-peer metrology and 3-dimensional mapping of relative locations

231 ) shipping container

232) identified informative points on a physical structure equipped with a transponder for peer-to-peer metrology and 3-dimensional mapping of relative locations

502) Plan inspection 504) Conduct inspection 506) Do inspection results suggest a need for for the EMR? 508) End the inspection, file the report.

510) As part of the inspection, deploy the electromagnetic climbing robot (EMR) to evaluate details not easily accessed by other means.

512) Are the EMR results sufficient for the purposes of the inspection?

514) End the inspection, file the report.

516) Determine any additional tools or supplies required by the EMR.

518) Launch the protected VTOL to carry said additional tools or supplies to the EMR.

520) Do the inspection results indicate a need for further EMR work? 522) End the inspection, file the report.

540) Probability distribution of possible conditions for each structure in an inventory

542) New data

544) Revise probability distributions

546) Plan inspections and maintenance for the inventory

548) Event of interest, e.g., further inspection or maintenance

550) Update probability distribution to consider the event of interest

602) Model of the knowledge and uncertainty of the state at time / given information available at time v expressed as a probability distribution D _t \ _v .

604) Latest observation. 606) Indicator of the plausibility of y, given D,\ _v .

608) Evaluation of whether D,\ _v still provides plausible forecasts fory, or should be updated.

610) Updating the previous reference distribution D, _\v to v = /.

DETAILED DESCRIPTION - FIRST EMBODIMENT: DYNAMIC DEFLECTION METROLOGY - FIGS IA, I B, 1C

One embodiment of our novel dynamic deflection metrology is depicted in Figs I A, I B, and 1C. As indicated in Fig IA, a span of a bridge 101 will be provided with a stable reference point 104 near an identified informative point 103, typically in the middle longitudinally of a span. Any vertical deflection of the bridge deck 102 will translate into a change in the distance between the identified informative point 103 and the stable reference point 104. The nominal stability of the stable reference point 104 will be achieved by connections 106 of a fixed length between the stable reference point 104 and two normally stable points 105 on the structure or environment 107, such as piers or abutments of a bridge. In some cases, the length of the constant length of connections 106 shall be achieved by a spring (not shown) whose tension minimizes the distance between the identified informative point 103 and the stable reference point 104. In such cases, the connections 106 might be provided by cables.

In some cases, the means to measure the distance between the identified informative point 103 and the stable reference point 104 will be supported by a mounting bracket 111 affixed to the under side of the bridge deck 102, as depicted in Fig I B. Linear displacement or position sensor(s) will be held in fixed position(s) relative to the identified informative point 103 and other portions of the bridge deck 102 in close proximity to the identified informative point 103. This embodiment may further include a reference assembly such as the telescoping tubes sketched in Fig 1C. The reference assembly may use something other than telescoping tubes, but "telescoping tubes" will be described here for ease of exposition. The inner tube 121 of the telescoping tubes will be affixed via, e.g., a ball joint to the identified informative point 103. The outer tube 122 of the telescoping tubes will essentially provide the stable reference point 104 of Figs I A, I B and 1C. Vertical deflections will then be measured by a linear displacement or position sensor affixed to the mounting bracket 111 at a mount 123. The line 124 in Fig 1 C indicates either the linear displacement or position sensor or something connecting the outer tube 122 to the linear displacement or position sensor affixed to the mounting bracket 123.

In some installations, there may be concern about excessive horizontal motion of the telescoping tube assembly, creating non negligible horizontal components to changes in the distance measured by the linear displacement or position sensors affixed to the mounting bracket 123. To provide a means to correct for such problems, the telescoping tube assembly may be provided optionally with a second linear displacement or position sensor 126 affixed to a second mount 125, whose location is selected to support measurement of a change in the horizontal position of the outer tube 122 relative to the mounting bracket 111. In such cases, this horizontal measurement may be replaced or refined by a direct measurement of the distance between the nominally stable points on the structure or environment 105 of Figs I A and 1 B using another linear displacement or position sensor; this other linear displacement or position sensor is not indicated in the Figs.

OPERATION - FIRST EMBODIMENT: DYNAMIC DEFLECTION

METROLOGY - FIGS I A, I B, 1C

Any physical device or structure bends or deflects as the load changes. As the vertical load on the bridge deck changes, the vertical distance between the identified informative point 103 changes relative to the stable reference point 104. Similarly, a change in the torsional load on the bridge will effect a change in the relative positions of the identified informative point 103 and the stable reference point 104, unless they are located in the lateral center of the bridge. Such changes will be measured by a linear displacement or position sensor 124. If a horizontal displacement of the reference assembly 121, 122, providing the stable reference point 104, is sufficient to produce an unacceptable discrepancy between the change in vertical distance and the length of the vertical linear displacement or position sensor 124, a second linear displacement or position sensor 126 may be installed to support simultaneous evaluation of vertical and horizontal motion of the reference assembly. Trilateration, using the theorem of Pythagoras, can be used to translate the horizontal and vertical distances between the mounts 123 and 125 and the hypotenuse information from the linear displacement or position sensors 124 and 126 into the desired changes in vertical deflection of the bridge and horizontal displacement of the reference assembly 122.

DETAILED DESCRIPTION - SECOND EMBODIMENT: DYNAMIC REMOTE OBSERVATION USING REFLECTIVE MEANS

A second embodiment places reflective means at identified informative points on structures of various kinds. Fig 2A shows a dam 201 with reflective means installed at one or more identified informative points 202. In some installations, the precise locations of the identified informative points 202 might be measured by monitoring devices 203 installed in positions presumed to remain fixed.

Fig 2B depicts a bridge 211 with reflective means 202 marking the longitudinal mid span of the bridge on both sides of the bridge laterally. The positions of the two reflective means 202 are monitored by at least one monitoring device 203 mounted to the abutments of the bridge. A marine oil or gas drilling or productions platform 221 is depicted in Fig 2C, whereon a plurality of identified informative points 222 are each provided with a transponder for peer-to-peer metrology and 3-dimensional mapping of relative locations.

Monitoring for changes in the physical shape of a container 231 such as a mobile hazmat container is outlined in Fig 2D. In this case, a plurality of transponders 232 communicate either electromagnetically or acoustically or both, possibly in multiple bands in either or both electromagnetics and acoustics, to establish their relative locations and possibly to simultaneously provide other information about the contents of the container.

Figs 2A, 2B, 2C and 2D illustrate only a few of the many types of structures that could profitably be monitored via reflective means installed on a plurality of informative points on physical structures of almost any type, including buildings, cliff faces, towers and large sculptures, to name only a few.

OPERATION - SECOND EMBODIMENT: DYNAMIC REMOTE

OBSERVATION USING REFLECTIVE MEANS

The reflective means in Figs 2A and 2B could be either passive like corner reflectors or active like transponders of various kinds. The locations, including any changes, would be measured by monitoring devices installed elsewhere on the structure or in the surroundings. The monitoring devices can be equipment to measure one, two or all three of range, elevation and azimuth. A three-dimensional map of the locations of the identified informative points 202 relative to each other and to the monitoring devices 203 will then be constructed using some appropriate optimization algorithm combining techniques as appropriate from the available literature on trilateration, triangulation and metric multidimensional scaling. Such measurements can be taken using either acoustics or electromagnetic radiation of various frequencies. In some installations, the reflective means could be essentially the same equipment as the monitoring devices, arranged in a peer-to-peer network. In some installations, the monitoring devices could be video cameras with a light, possibly operating outside the visible spectrum, illuminating the corner reflectors making them stand out very clearly from other features in the scene. The use of corner reflectors solves arguably the most difficult problem of photogrammetry, namely precise identification of the locations to track. This permits a substantial improvement in the precision otherwise obtainable from video camera technology.

In installations similar to those depicted in Figs 2C and 2D, the reflective means would be transponders in a peer-to-peer arrangement, at least one of which communicates with the outside world. Each of the transponders at identified informative points (222 in Fig 2C, 232 in Fig 2D) measures its distance from some but not necessarily all of the other transponders. The result is then used to construct a three-dimensional map of the relative locations of the identified informative points, 222 or 232, using techniques such as metric multidimensional scaling.

DETAILED DESCRIPTION - THIRD EMBODIMENT: PROTECTED VTOL

A VTOL (vertical take-off and landing) helicopter will be provided with protective bumper bar(s) (provided by one or more protective rods 1) to allow it to safely approach and physically contact a structure in a mode other than landing on a horizontal surface. This will support taking pictures of higher resolution than feasible otherwise as well as the installation of measurement devices and associated communications equipment for long term monitoring of the condition of the structure. Images taken might be in any portion of the electromagnetic spectrum deemed to provide useful information about the condition of the structure and may in some cases be stereoscopic depending on the specific imaging equipment (e.g, television camera) carried by the VTOL device. The VTOL device could also carry remote sensing and communications equipment that could be affixed to the structure with glue or magnets or some other system deemed to be sufficiently safe and permanent.

It could also make deliveries to a person or a robot in location with limited accessibility; such deliveries could include tools, paint, or other equipment or supplies to be installed. Sensors installed might include but not be limited to acoustic metrology, inclinometers or accelerometers that could detect seismic events or accidents where, for example, a boat, barge, truck, train, or object protruding from any of these might strike a structural element with such force as to jeopardize the physical strength of the structure.

Multiple devices installed could potentially also measure their distances from one another, as described in Figs 2C and 2D. The structure sensors could also be augmented with weather sensors of various types indicating atmospheric conditions before, during and after any data events collected by the actual structure sensors. These weather sensors include but are not limited to: anemometers, wind veins, temperature, humidity, barometric, Doppler radar, and Lidar.

The Protective Rod(s), (Bumper Bars) 1 are constructed of thin, light weight material such as fiberglass rod, formed spring steel or similar. The material should exhibit slight flexibility to absorb impact but sufficient rigidity to avoid contact with the rotor blades 2 when stressed from contact. The structure of rods (Bumper Bars) mounts to the VTOL device at the base skids, landing gear or undercarriage of the fuselage 4. The structure may have as few as one protective rod (Bumper Bar) to a maximum of as many that can be carried by the VTOL and still maintain the aerodynamic flow required for lift. The structure may be completely open at top and unconnected or may be connected to other protective rods (Bumper Bars) in any appropriate manner which can still allow the lifting attribute of the rotors to function properly. In many cases fine wire such as .032 safety wire or smaller or simple fishing line is affixed to the protective rods at the points possibly even with the rotor blades. This provides both additional rigidity and additional side protection against intrusion by guy wires or other thin structure.

OPERATION - THIRD EMBODIMENT: PROTECTED VTOL

As a fixed and attached system of the helicopter, one or more curved protective rods (Bumper Bars) 1 form an extended bumper from the body, base or skids (fuselage 4) of the helicopter which reduces the chance that the helicopter will touch a fixed object with the rotor blades 2 or tail rotor 3. Only one protective rod (Bumper Bar) may be used to protect a VTOL aircraft such as where the tail rotor 3 is. This could possibly be where there is ample forward and side vision but the craft may not be equipped with the capability to have operator knowledge of the obstacle clearance of the tail rotor. Another application of a single protective rod (Bumper Bar) 1 might be where the VTOL device needs to move very close to an object and will only do so in the forward position. All other operations are to be through directly approaching the object, in which case only one protective rod (Bumper Bar) 1 is required on the front of the vehicle.

However, when the power to weight ratio is ample enough to allow the designer to include numerous protective rods (Bumper Bars) 1 an optimum number could be sixteen or more.

DETAILED DESCRIPTION - FOURTH EMBODIMENT: CLIMBING ROBOT, F1GS 4A, 4B, 4C

Two or more feet of an existing climbing robot can be replaced with the electromagnetic feet outlined with Figs 4A, 4B, 4C, thereby making it an EMR (Electromagnetic Robot). The base robot modified in this way could be of any design otherwise suitable. Our initial candidate might be a device of the "RiSE" robot project (2009) or something similar. An alternative might be something using "sticky foot" technology (Graham-Rowe 2007), with some or all of the sticky feet replaced by the magnetic feet of Fig 4A, 4B, 4C. The sizes of EMRs could range from quite small and inexpensive to much larger.

An EMR is able to maneuver itself via remote control (wire or wire-less) or programmed control to a specific location and perform work in transit or at a location of interest. This work may be in the form of optical observation, clearing or scraping an area, sensing or installing sensors. The smallest EMRs might climb to a remote location and remain in place collecting data and relaying it to a central location using any of several communications protocols including, for example, wireless local area network technology. Larger EMRs might climb to various locations to install smaller data collection and communications devices in appropriate places using a variety of sensing technologies capable of providing information from which an assessment of the condition of the structure, often in combination with data from other portions of the structure and / or data of other types.

The magnetic feet can allow a climbing robot to climb steel structures where flaking paint and rust may make it infeasible to use existing "sticky foot" technology (e.g., Graham-Rowe 2007) and the surface may not be sufficiently rough for the RiSE (2009) gripper feet.

Magnetic feet could be used to move horizontally or at a slight grade even if the surface had little or no ferromagnetic content.

Some EMR configurations might replace only some of the feet of a climbing robot, thereby allowing the unit to traverse mixed terrain, where for example gripper feet might work for part of the distance to the destination without adequate ferromagnetic content to allow the magnetic feet to work, and the magnetic feet would work where gripper feet might not have sufficient grip. Units designed to be left in place might also carry other adhesives to glue themselves to the structure after arriving at their desired permanent location.

EMR robots could perform tests or install equipment requiring more solid contact with the structure than is feasible with the VTOL device. In some configurations, they could accept deliveries of supplies and equipment from a VTOL. This would substantially reduce the need for a relatively slow-moving device to interrupt work and return to the ground for more supplies required to continue work. This would allow an EMR to do more work in less time than might otherwise be feasible.

OPERATION -- FOURTH EMBODIMENT: CLIMBING ROBOT, FIGS 4A, 4B,

4C

In motion, the robot 28 moves a leg or arm 27 to position the magnetic foot 21 for placement. The permanent magnet 25 of the foot 21 extends to slightly beyond the non-magnetic foot pad/cushion 26. As the foot nears the surface, the magnetic force between the permanent magnet and the ferrous material of the structure form a magnetic bond and hold the foot and surface until other action force is applied.

To take another step, the electromagnet shell 24 would be energized to push the permanent magnet 25 away from the ferromagnetic surface by solenoid action. This then allows it to be moved to a new location. No power from the robot is required for maintenance of this attachment (connection) for as long as the built in permanent magnet retains its magnetic field. Thus the robotic device with this magnetic foot can operate for very long periods of time in any position in which it comes to rest without consuming power to remain in place.

Each foot that the robot has in its design will function similarly. If the robot is equipped with only two legs or arms then one will maintain hold on its' fixed position while the other repositions. If the robot is equipped with three or more legs or arms, then the designer will be able to determine if only one leg or arm repositions at a time or if two may reposition at the same time. For example, a robot with six legs or arms may have the ability to use two to maintain a hold of its position while the other four legs or arms simultaneously release and reposition.

The functions of each specific part of the magnetic foot are as follows: The cap 22 closes the electromagnetic coil and provides a fixed backing for the positioning spring 23. The positioning spring returns the permanent magnet 25 to it's extended location so that it is ready to attach to a surface. The electromagnet shell 24 provides a cylinder for the permanent magnet to travel within and contains an electromagnetic coil controlled by the robot to release the permanent magnet 25 from the surface to which it is connected. The nonmagnetic foot pad / cushion 26 provides a buffer zone for the electromagnet so that the electromotive force generated to remove the permanent magnet is isolated from the surface to keep the electromagnet shell 24 from also adhering to the surface; it provides a small amount of shock absorption to the robot when the permanent magnet 25 initially contacts a surface.

DETAILED DESCRIPTION - FIFTH EMBODIMENT: DATA COLLECTION AND ANALYSIS SYSTEM, FIGS 5A, 5B, 5C

As indicated in Fig 5A, the protected VTOL device would be deployed whenever appropriate to inspect a structure of interest 502 using metrology equipment such as a video camera, recording the results 504. If the data collected thereby seemed sufficient for the purposes 506, the inspection would end 508. Otherwise, as indicated in Fig 5B, some other device such as the EMR described above or a human 510 would be deployed to more closely examine the remote area of interest, collecting data and possibly installing remote metrology equipment that might reduce the need for future periodic inspections by equipment such as the protected VTOL device. In some situations, the human or other device 512 may need additional supplies and equipment 514 that can then be delivered using the VTOL 516. Upon completion, the VTOL can reinspect the areas of concern and / or return to base 518. If the inspection is complete 520, the process ends 522. Otherwise, it may require another trip with the VTOL to carry more supplies to the EMR or human. Using the VTOL and EMR in this way would substantially reduce the time and money required for certain remote maintenance tasks.

Of course, many modifications on this basic process are possible and may be obvious and profitably employed by someone skilled in the art.

OPERATION - FIFTH EMBODIMENT: DATA COLLECTION AND ANALYSIS SYSTEM, FIGS 5A, 5B, 5C

Data collection including preexisting data, data collected during site visitation, data collected over time as constant streaming data and/or data from specific time periods or triggered events is sent through electronic means e.g., wireless and web based systems, to a central data collection repository. This data is then accessible as a continuous input stream and/or an accessed file for analysis.

The use of data collected by remote access devices such as the protected VTOL or EMR or by metrology units installed using them is new. In operation, such data collected might be used as follows:

Step 0 (540). The structure monitoring process is initiated by accessing potentially a variety of available data sources such as the National Bridge Inventory, subjective judgments by knowledgeable personnel and other sources deemed relevant. This information is converted into an assessment of the health of each structure (e.g., bridge) of interest at that time. This could be a qualitative summary or a more formally specified probability distribution over some list of possible alternative states of health or deterioration. This could optionally be performed using fuzzy theory, logistic regression, or data mining techniques; many others are known in the literature and available in widely available software. If a probability distribution is used, it will allow users to compute for each structure a probability of failure in a certain period of time, e.g, 1 or 5 years, under both normal conditions and stress conditions such as flood, storm surge, high wind, and earthquake. These numbers can be combined with traffic, bridge length and perhaps other information to estimate the expected loss of life due to bridge failure in a particular planning horizon. For example, suppose the probability of failure for Bridge A within the next 5 years is estimated at 0.1 or 10%. Suppose Bridge A is short and remote and has only 10 vehicles crossing it each day, and a heuristic or more formal model we develop might estimate an average of only 0.1 vehicle carrying one person would be on that bridge at the time of failure with only a 2% chance of fatality if the bridge fails with a vehicle on it. We could combine these in some reasonable way to estimate 10% of 0.1 times 2% gives us an average of 0.0002 lives from bridge failure on that bridge over the next 5 years. Suppose that Bridge B is one kilometer long, carrying on average 200 vehicles every second of every day. And suppose that it has a 1% chance of failure in the next 5 years. However, suppose it's taller and it's design is not sufficiently redundant, so that we would estimate 1.01 fatalities in 10% of the vehicles on the bridge at the time of failure, giving us an average of 1% of 200 times 1.01 times 10% = 0.202 fatalities. If the cost of further data collection to improve our estimates of the probabilities of failure, etc., are not grossly different between the two bridges, we would want to inspect Bridge B first because it has a substantially larger estimated loss of life from bridge failure in the next 5 years. These probabilities of failure and loss of life may be excessive for most bridges, but we will use them here for the sake of illustration. Other procedures could be used to produce an evaluation of the condition of each structure in the inventory of a particular responsible manager. This evaluation could be pure qualitative or totally quantitative as just described or some mixture, possibly developed by a procedure such as the Analytic Hierarchy Process (Wikipedia, "Analytic Hierarchy Process").

Step 1 (544). For each planning cycle and administrative decision, we upgrade the evaluation of structural adequacy to consider data acquired since the last evaluation. If this evaluation includes probabilities, then for example it might make the new estimate of the probabilities of failure 1 1 % for Bridge A and 1.2% for Bridge B due to differences in usage patterns, for example. And suppose that the traffic on Bridge A has not changed while that on Bridge B has increased by 10% to 220 vehicles on average. Then the new estimated fatalities might be 0.00022 for Bridge A and 0.266 for Bridge B. Similar computations for all other structures in a particular inventory could be used to help decision makers allocate resources for bridge inspection and maintenance using a variety of procedures involving different mixtures of heuristics and mathematical formalisms.

Step 2 (546). Decision makers then plan further inspection and maintenance activities based on part on the Step 1 evaluations.

Step 3 (548). An event of interest occurs. This could be an inspection or maintenance activity planned in Step 2. It could involve the permanent installation of real-time data collection systems on bridge condition, in which case, each discrete report from such a system might itself be an event of interest in the future.

Step 4 (550). At regular intervals or at times triggered by events of interest or both, the evaluations of Step 1 (544) are repeated. Parts of this evaluation could be automated. If the assessment exceeds some threshold, it could trigger some action. The action could be an email sent to an appropriate official or even a robot phone call with a message briefly describing the condition and the contingency plan to be implemented. It could be a signal from a central computer based in part on weather conditions or seismic activity sufficiently close to a particular problem structure that raise the probability of an impending collapse to an unacceptable level. This signal might go either to local emergency services or to traffic signals installed on each end of the bridge that might not be visible under ordinary circumstances but could be deployed remotely when the data collected suggested the need. If the structure monitored is a railroad bridge, this action could go directly to the railroad signal system, suspending traffic over that bridge until an engineer completes a more careful review. In essence, the bridge could call "91 1 ". DETAILED DESCRIPTION - SIXTH EMBODIMENT: TIME SERIES DATA COMPRESSION, FIG 6

Fig 6 outlines a standard Bayesian sequential updating process with an important difference that the state is not updated with each observation but only when recent observation(s) are implausible and updating the state would produce more accurate predictions.

For the variation in Fig 6, let D,\ _v = the information available at time v about the condition of the system being monitored at time /, 602 in Fig 6. In the worst case D,\ _v consists of all the data stored up to time v used to forecast the value and uncertainty in the forecast at time /. Using classical Bayesian statistics (e.g., Graves 2007; Graves et al. 2001 ), this will typically be summarized as a multivariate normal distribution over some state space representation of the process, though fuzzy theory could also be used. There is a substantial literature on system identification that can be quite valuable for selecting an appropriate state space representation to efficiently summarize the condition of the physical structure being monitored in some cases. However, for many applications, it may suffice to compress the data from each sensor individually.

Arguably the simplest state space model is an exponentially weighted moving average. This produces an autocorrelation structure among observations^, comparable to a first order moving average on the first differences, IMA(I ₅ I), of the observations^, 604. In this case, there is a known relationship between the IMA parameter and the noise and migration variances in the corresponding state space model.

For physical parameters like temperature and perhaps deflection, a two- or three-dimensional state space representation may be more appropriate, with the three dimensions being mean, velocity or time rate of change in the nominal level, and optionally acceleration. For measurements such as temperature, it may be desirable to have the two dimensions represent the mean and velocity of deviations from some model predicting the temperature being read as a function of data from other sources such as ambient temperature at a nearby national weather service station.

In some cases, D,\ _v may consist of a mixture of multivariate normal

distributions, one representing the "good" condition of the structure or sensor, while others represent various modes or states of deterioration, with the mixture

probabilities being an efficient summary of the chances that the alternative states are the most accurate description of the existing condition.

An observation ^, could be anything, univariate or multivariate, continuous, discrete counts or categorical. The key point is that it can be compared with D _t \ _v to produce an index of plausibility or consistency p,\ _v of the current observation with D,\ _v . With normally distributed observations linearly related to the state with normal uncertainty, this might be expressed as the probability of obtaining an outcome at least as extreme as that observed.

With observations^, that are usually normally distributed but may occasionally exhibit outliers, it may make sense to ignore one outlier but not several in succession. Thus, the decision regarding plausibility may be based on examining p,, _\v , u = v, v+1 , ... ( in a way tailored to the specific nature of the sensors being used.

If the recent data are deemed to be inconsistent with D,\ _v , the reference state is updated so v = /, and latest state is updated to D,\, and is then stored in the database for future reference. On a structure employing smart sensors, this data compression may occur at the sensor level before being reported to the central computer controlling all the sensors on the structure or on a certain portion of the structure. These may further be relayed to the central database unchanged. Or they may be further compressed using a larger state space representation for the entire structure, recursing on this algorithm, before reports are sent to the central database.

In some cases, observation(s) y, may be inconsistent with both D ,\ _v and - In such cases, the algorithm could report at least a subset of such outlying y,'s or switch to some other data compression scheme that would provided better diagnostics to help identify the sources of such discrepancies. For example, a sensor could be malfunctioning. Alternatively, the structure could be behaving in ways inconsistent with the state space representation assumed by D,\ _v .

In any event, more data and more state summaries may be stored locally than are transmitted to the central database. In this way, the local computers and smart sensors can act similar to the flight recorders on aircraft, storing fine details from recent data to facilitate forensic engineering analyses of any structural collapse. Such a system could provide extremely valuable detailed information about the collapse process not obtainable in any other way.

OPERATION -- SIXTH EMBODIMENT: TIME SERIES DATA COMPRESSION, FIG 6

One incarnation of the core data compression algorithm proposed here is summarized in Fig 6, though variations are possible. For the variation in Fig 6, let D,\ _v = the information available at time v about the condition of the system being monitored at time /, 602 in Fig 6. New data y, 604 are received at time /.

An index of plausibility p,\ _v 606 is computed quantifying the level of consistency between^, and D,\ _v . Some assessment is made of the plausibility of the data collected between v and t, p _u \ _v , u = v, v+1 , ..., / 608. If the new data are consistent with D,\ _v , nothing is done other than wait for the next observation. If the new data are not consistent with A _| v, v is replaced by /, and the currently available information is stored as A _| v 610. CONCLUSION, RAMIFICATIONS AND SCOPE

Deficiencies in current practices in structural health monitoring were outlined above in discussing the prior art regarding current bridge inspection practices. Other deficiencies have been exposed by the recent Deepwater Horizon accident (Wikipedia "Deepwater Horizon oil spill").

The current patent application proposes two types of improvements on these practices. Perhaps most importantly, the machines and processes taught herein provide better quality information regarding structural health at a lower cost than alternatives currently available. Current practices focus primarily on visual inspections and analyses of vibrations. Visual inspections often miss important problems, and analyses of vibrations are extremely difficult, requiring detailed models of the structure that often may not adequately reflect the current deteriorated state of the structure.

The dynamic deflection metrology proposed herein provides novel and valuable means for directly measuring deflection in response to changing load, arguably the single most important indicator of structural health. It has largely been overlooked in the prior art, we believe, because it was so difficult that it became unthinkable. A method is also taught herein for optimizing data compression and storage designed using novel considerations of both physics and metrology noise.

Other portions of the present application provide means for approaching otherwise inaccessible structures safely. One use would be to take higher resolution pictures than is otherwise economically feasible with current practices. Other uses could be to perform other metrology tasks or to install small metrology units for longer term monitoring. REFERENCES CITED:

U.S. PATENT DOCUMENTS:

number issued inventor(s)

2,966,797 1/1961 Brewer et al. .

4,480,480 1 1/1984 Scott et al. .

4,698,634 10/1987 Along, et al. .

4,789,947 12/1988 Maciejczak .

4,843,372 6/1989 Savino .

5,167,295 12/1992 Moog .

,404, 132 4/1995 Canty et al. .

,421,204 6/1995 Svaty .

,507,188 4/1996 Svaty .

,698,787 12/1997 Parzuchowski et al. ,857,534 6/1999 DeVault .

,012,337 1/2000 Hodge .

,023,061 2/2000 Bodkin .

,192,758 2/2001 Huang .

, 181,841 1/2001 Hodge .

,240,783 6/2001 McGugin . 6,351,218 2/2002 Smith.

6,459,760 10/2002 D'Ambrosio .

6,476,377 11/2002 Hodge .

6,487,914 12/2002 Hodge.

6,581,466 6/2003 Costleyetal.

6,647,161 11/2003 Hodge.

6,703,600 3/2004 Hodge .

6,722,202 4/2004 Kennedy .

6,646,559 11/2004 Smith .

6,862,528 3/2005 Scannell .

6,947,842 9/2005 Smith et al..

7,044,074 5/2006 Harrie .

7,394,361 7/2008 Twitchell .

7,510,142 3/2009 Johnson.

7,554,442 6/2009 Twitchell .

7,572,360 8/2009 Moshier et al.

7,733,818 6/2010 Twitchell.

7.742.772 6/2010 Twitchell.

7.742.773 6/2010 Twitchell. U.S. PATENT APPLICATIONS:

number filed inventor(s)

2002/0080038 6/2002 Smith . 2006/0226280 10/2006 Alman . REFERENCES

Aktan, A. Emin, Catbas, F. Necati, Grimmelsman, Kirk A., Pervizpour, Mesut (2003) Development of a Model Health Monitoring Guide for Major Bridges (Drexel Intelligent Infrastructure and Transportation Safety Institute, Drexel U.;

www.di3.drexel.edu/w2/files/FHWA_Report_7_l 8_03.pdf, accessed

2010.06.30.)

Alman, James (2006) "Stowable Design for Unmanned Aerial Vehicle", US Patent Application No. 2006/0226280, Oct. 10.

Alongi, Anthony V., and Alongi, Anthony J. (1987) "Subsurface Inspection Radar", US patent number 4,698,634, Oct. 6.

Azinheira, Jose Raul, and Moutinho, Alexandra (2008) "Hover control of a UAV with back stepping design including input saturation", IEEE Transactions on Control Systems Technology, 16(3): 517-516.

Benjamini, Yoav, and Hochberg, Yosef (1995). "Controlling the false discovery rate: a practical and powerful approach to multiple testing" Journal of the Royal Statistical Society, Series B (Methodological) 57 ( 1 ): 289-300.

Biezma, Maria Victoria, and Schanack, Frank (2007) "Collapse of steel bridges", Journal of Performance of Constructed Facilities, 21 (5): 398-405. Blanchard, Steve (2003) "Guidance for Bridge Scour Studies" (US Geological Survey, http://water.usgs.gov/admin/memo/SW/sw03.06.html, accessed 2009.01.04).

Bodkin, William Andrew (2000) "Miniature Infrared Camera", US patent number 6,023,061 , Feb. 8.

Boston Dynamics (2010) "What Robots Can Do" (www.bostondynamics.com,

accessed 2010.06.30).

Box, George, Graves, Spencer, Bisgaard, Søren, Van Gilder, John, Marko, Ken,

James, John, Seifer, Mark, Poublon, Mark, and Fodale, Frank (2000) Detecting Malfunctions in Dynamic Systems (Society of Automotive Engineers Document number 2000-01 -0363, "www.sae.org/technical/papers/2000-01 -0363", accessed 2009-08-01)

Brehm, Denise, (2007) "Handheld device 'sees' damage in concrete bridges, piers", MIT News, May 18, 2007; "http://web.mit.edu/newsoffice/2007/inspection- 0518.html", accessed 2008.12.28.

Brewer, Ward B., Brodner, Martin, and Smith, Robert H. (1961 ) "Electrical Gage for Measuring Mechanical Deflections", US patent number 2,966,797.

Canty, Jeffrey N., and Canty, Charles W. (1995) "Deflection Monitoring System", US patent number 5,404,132

Carden, E. Peter, and Fanning, Paul (2004) "Vibration Based Condition Monitoring:

A Review", Structural Health Monitoring, 3(4): 355-377.

Catbas, F. N., Ciloglu, S. K., Hasancebi, O., Grimmelsman, K., and Aktan, A. E.

(2007) "Limitations of Structural Identification of Large Constructed

Structures", Journal of Structural Engineering, 133(8): 1051 - 1066. Chang, Peter C, Flatau, Alison, and Liu, S. C. (2003) "Review Paper: Health

Monitoring of Civil Infrastructure", Structural Health Monitoring, 2(3): 257- 267.

Cheng, Jimming, and Metzenmacher, Michael (2005) "The Markov Expert for

Finding Episodes in Time Series", Proceedings of the 2005 Data Compression Conference (wvvw.eecs.harvard.edu/~michaelm/postscripts/jsubmit.pdf, accessed 2010.07.01).

Choi, Myoungkeun, and Sweetman, Bart (2009) "Efficient Calculation of Statistical Moments for Structural Health Monitoring", Structural Health Monitoring, 9: 13-24.

Costley, R. Daniel, Henderson, Mark E., and Dion, Gary N. (2003) "Acoustic

Inspection of Structures", US patent number 6,581 ,466, June 24.

D'Ambrosio, Karl V. (2002) "Apparatuses and Methods for Non-destructive

Inspection", US patent number 6,459,760, Oct. 1.

Darwin, David, Browing, JoAnn, and Lindquist, Will D. (2004) "Control of Cracking in Bridge Decks: Observations from the Field", Cement, Concrete and

Aggregates, 26(2) 148: 154.

De Vault, James E., Hudson, William B., and Hussain, Mustaque (1999) "Robotic Inspection Apparatus and Method", US patent number 5,857,534, June 12.

Farhey, Daniel N. (2006) "Instrumentation System Performance for Long-term Bridge Health Monitoring", Structural Health Monitoring, 5(2): 143- 153.

FHWA: United States Federal Highway Administration (2004) National Bridge Inspection Standards (NBIS, www.fhwa.dot.gov/bridge/nbis.htm, accessed 2010.06.14). Fink, Eugene, and Gandhi, Harith Suman (2007) "Important Extrema of Time Series", Computer Science, Carnegie Mellon U.

(www.cs.cmu.edu/~eugene/research/full/ important-extrema.pdf, accessed 2010.07.01).

Glisic, Branko, and Inaudi, Daniele (2007) Fiber Optic Methods in Structural Health Monitoring (Wiley).

Gordon, Robert, and Knopf, Robert (2005) "Evaluation of Wrought Iron for Continued Service in Historic Bridges", Journal of Materials in Civil Engineering, 17(4)393-399.

Graham-Rowe, Duncan (2007) "Wall-Climbing Robot: A newly created robot

improves upon a gecko's sticking power." Technology Review, April 30, 2007. "www.technologyreview.com/Infotech/18602", accessed 2009.05.16.

Graves, Spencer (2007) "Bayes' Rule of Information and Monitoring in

Manufacturing Integrated Circuits", Blanca Colosimo and Enrique del Castillo (2007) Bayesian Process Monitoring, Control and Optimization (Chapman & Hall)

Graves, Spencer, Bisgaard, Søren, Kulachi, Murat, Van Gilder, John, Ting, Tom, Marko, Ken, James, John, Zatorski, Hal, and Wu, Cuiping (2001) Foundations of Monitoring Dynamic Systems (www.prodsyse.com; accessed 2009.08.02).

GuI, Mustafa, and Catbas, Necati (2008) "Ambient Vibration Data Analysis for

Structural Identification and Global Condition Assessment", Journal of

Engineering Mechanics, 134: 650-662.

Harrie, Per, Fagerholm, Johan, Kurdve, Michael, Kindsjo, Johan (2006) "Method and Arrangement for Inspection of an Object", US Patent No. 7,044,074, May 16, 2006. Hart, Ariel (2008) "DOT employee falsified bridge inspection reports", Atlanta Journal-Constitution, Feb. 4, 2008

(http://www.ajc.com/metro/content/metro/stories/2008/ 02/04/DOT0204.html: accessed 2008.12.28).

Hodge, Malcolm H. (2000) "Structural monitoring sensor system", US Patent Number 6,012,337, Jan. 1 1. (2001 ) "Structural monitoring sensor system", US Patent Number 6, 181 ,841 ,

Jan. 30. (2002a) "Structural monitoring system", US Patent Number 6,476,377, Nov.

5.

(2002b) "Structural monitoring sensor system", US Patent Number

6,487,914, Dec. 3.

__ (2003) "Structural Monitoring Sensor System", US Patent Number 6,647,161 , Nov. 1 1. (2004) "Structural monitoring apparatus", US Patent Number 6,703,600,

Mar. 9.

Huang, Kang (2001 ) "Structure Safety Inspection", US Patent Number 6,192,758, Feb. 27.

Humar, Jag, Bagchi, Ashutosh, and Xu, Hongpo (2006) "Performance of Vibration- based Techniques for the Identification of Structural Damage", Structural Health Monitoring, 5(3): 215-241.

Izumi, Tetsuya, and Iiguni, Youji (2006) "Data Compression of Nonlinear Time

Series using a Hybrid Linear / Nonlineaer Predictor", Signal Processing, 86: 2439-2446. Johnson, Samuel A. (2009) "Aerial Robot", US Patent Number 7,510, 142, Mar. 31.

Kennedy, James C, Mares, Christopher L., and Negley, Mark A. (2004) "Method and Apparatus for Inspecting a Structure Utilizing Magnetically Attracted Probes", US patent number 6,722,202, Apr. 20.

Lagasse, P. F., Arneson, L. A., and Zevenbergen, L. W. (2006) "Research Needs for Scour at Bridge Foundations - An Update", in Randall Graham , ed.,

Proceedings of the World Environmental and Water Resources Congress 2006 (American Society of Civil Engineers, www.asce.org).

Lazaridis, Iosif, and Mehrotra, Sharad (2003) "Capturing Sensor-Generated Time Series with Quality Guarantees", Proceedings of the 19th International Conference on Data Engineering, 429-440).

Li, Lei (2010) "Fast Algorithms for Time Series Mining", ICDE Workshops 2010, 341 -344.

Livingston, Richard A., Mills, Milton, and Oskard, Morton S. (2002) "A decade of achievement", Public Roads, Nov. / Dec. 2002: 38-42

(http://www.fhwa.dot.gov/ publications/publicroads/02nov/09.cfm, accessed 2010.06.30).

Lopez, M. Rivas , Sergiyenko, O. Yu, Tyrsa, V. V., Perdomo, W. Hernandez, Cruz, L. F. Devia, Balbuena, D. Hernandez, Burtseva, L. P., and Hipolito, J.I. Nieto (2010) "Optoelectronic Method for Structural Health Monitoring", Structural Health Monitoring 2010; 9; 105- 1 16.

Maciejczak, Robert A. (1988) "Inspection System and Method for Mechanical

Structures", US patent number 4,789,947, Dec. 6.

Matech (2009) "The Metal Fatigue Problem" (http://www.matechcorp.com/metal- fatigue.html, accessed 2009.06.12)

McGugin, Terry Curtis, Wiggs, George Ashley, and Bishop, Stephen Keith (2001) "Bridge Monitoring System", US patent number 6,240,783, June 5.

Medwick, Seth, and Kaufman, Paul (2007) "Inspector Gadgets", Roads and Bridges, Nov. 2007: 66-69.

Metji, N., Chapman, D. N., Rogers, C. D. F., Henderson, P., and Beth, M. (2008) "An Optical Fiber Sensor System for Remote Displacement Monitoring of Structures ~ Prototype Tests in the Laboratory", Structural Health Monitoring, 7( 1 ): 51 - 63.

Metni, Najib, and Hamel, Tarek (2007) "A UAV for bridge inspection: Visual

servoing control law with orientation limits", Automation in Construction; 17(1): 3-10.

Moog, Alfons (1992) "Vehicle, in Particular a Chasis of a Bridge-underside

Inspection Apparatus", US patent number 5,167,295, Dec. 1.

Moore, Mark, Phares, Brent, Graybeal, Benjamin, Rolander, Dennis, and Washer,

Glenn (2001) Reliability of Visual Inspection for Highway Bridges (US Federal Highway Administration; "http://www.tfhrc.gOv/hnr20/nde/01020.htm", accessed 2008.12.28).

Moshier, Monty, and Berks, William I. (2009) "Electrochemical Fatigue Sensor

Systems and Methods", US Patent number 7,572,360.

Moustafa, Abbas, Mahadevan, Sankaran, Daigle, Matthew, and Biswas, Gautam

(2010) "Structural and sensor damage identification using the bond graph approach", Structural Control and Health Monitoring, 17: 178- 197.

Nagayama, T., Spencer, B. F., Jr., and Rice, Jennifer, A. (2009) "Autonomous decentralized structural health monitoring using sensors", Structural Control and Health Monitoring, 18: 842-859.

Nasif, Hani H., Gindy, Mayrai, and Davis, Joe (2005) "Comparison of laser Dopper vibrometer with contact sensors for monitoring bridge deflection and vibration", NDT & E International, 38: 213-218.

NCHRP: US National Cooperative Research Program (2005) "Inspection and

Management of Bridges with Fracture-Critical Details: A Synthesis of

Highway Practice", Transportation Research Board of the National Academies (http:// onlinepubs.trb.org/onlinepubs/nchrp/nchrp_syn_354.pdf, accessed 2010.06.290).

Olfati-Saber, Reza (2007) "Distributed Kalman Filtering for Sensor Networks," Proc. of the 46th IEEE Conference on Decision and Control

(http://engineering.dartmouth. edu/~Reza_Olfati_Saber/papers/cdc07_dkf.pdf, accessed 2010.06.04)

Overbey, L. A., and Todd, M. D. (2008) "Damage Assessment Using Generalized State-Space Correlation Features", Structural Health Monitoring, 7(4): 347- 363.

Parzuchowski, Halina M., Reighard, Mark K., and Kintz, Kevin G. (1997) "Portable Laser / Ultrasonic Method for Nondestructive Inspection of Complex

Structures", US Patent Number 5,698,787, Dec. 16.

Pozzi, Matteo, Zonta, Daniele, Wu, Huayong, and Inadi, Dnaiele (2009)

"Development and laboratory validation of in-line multiplexed low-coherence interferommetric sensors", Optical Fiber Technology, 14: 281-193.

Reddy, K. Ashoka, George, Boby, and Kumar, V. Jagadeesh (2009) "Use of Fourier Series Analysis for Motion Artifact Reduction and Data Compression of Photoplethysmographic Signals", IEEE Transactions on Instrumentation and Measurement, 58(5): 1706-171 1.

Rens, Kevin L., Nogueira, Carnot L., and Transue, David J. (2005) "Bridge Management and Nondestructive Evaluation", Journal of Performance of Constructed Facilities, 19(1) 3-16.

Rhazi, J. (2006) "Evaluation of Concrete Structures by the Acoustic Tomography Technique", Structural Health Monitoring, 5(4): 333-342.

"RiSE" robot project (2009) "to create a bio-inspired climbing robot with the unique ability to walk on land and climb on vertical surfaces."

(http://kodlab.seas.upenn. edu/~rise/newsite, accessed 2010.06.20).

Ryabko, Boris (2008) "Applications of Kolmogorov Complexity and Universal Codes to Nonparametric Estimation of Characteristics of Time Series", Fundamenta lnformaticae, 83 : 177- 196. (2009) "Compression-Based Methods for Nonparametric Prediction and

Estimation of some Characteristics of Time Series", IEEE Transactions on Information Theory, 55(9): 4309-4315.

Saadat, Soheil, Buckner, Gregory D., and Noori, Mohammad N. (2007) "Structural System Identification and Damage Detection using the Intelligent Parameter Varying Technique: An Experimental Study", Structural Health Monitoring, 6: 231-213.

Savino, Thomas (1989) "Bridge Sway and Deflection Detection System", US patent number 4,843,372,

Scannell, Joseph P. (2005) "Monitoring System and Process for Structural Instabilities Due to Environmental Processes", US Patent No. 6,862,528. Scott, David R., and Rhoades, Thomas S. (1984) "System for Assessing the lnntegrity of Structural Systems", US Patent No. 4,480,480.

Shevvhart, Walter A. (1931) Economic Control of Quality of Manufactured Product (1980 reprint: American Society for Quality Control).

Smith, Michael (2002) "Method and Apparatus for Activating Weather Warning Devices", US Patent No. 6,351 ,218.

Smith, Michael (2004) "Method and Apparatus for Activating Warning Devices", US Patent No. 6,646,559, Nov. 1 1.

Smith, Michael, and Long, Mark David (2005) "Normalized and Animated

Inundations Maps", US Patent 6,947,842 Sep. 20, 2005.

Sridhar, S., Ravisankar, K., Sreeshylam, P., Parivallal, S., Kesavan, K., and Murthy, S. G. N. (2009) "Development of a Real-time Remote Structural Monitoring Scheme for Civil Infrastructural Systems, Structural Health Monitoring, 8(6): 509-521.

Svaty, Karl J. (1995) "Structural Monitoring System", US patent number 5,421 ,204. (1996) "Structural Monitoring System", US patent number 5,507.188.

Thunder Tiger (2009) "Raptor helicopters", (www.thundertiger.com/products list/

18 1.html; see also "www.modelflight.corn.au/rc_rnodel_helicopters/ thunder_tiger_raptor_90_helicopter.htm", accessed 2009.06.12)

Titchener, M. R. (2008) "Towards Real-Time Measurement of Information in a

Scientific Setting", 6th International Symposium on Communication Systems, Networks and Digital Signal Processing, 316-320

Tobalske, Bret W., Warrick, Douglas R., Clark, Christopher J., Powers, Donald R., Hedrick, Tyson L., Hyder, Gabriel A., and Biewener, Andrew A. (2007) "Three-dimensional kinematics of hummingbird flight", Journal of

Experimental Biology, 210: 2368-2382.

TRB: US Transportation Research Board (1996) Scour at bridge foundations - research needs (National Cooperative Highway Research Program,

Transportation Research Board NCHRP 24-8; cited from Blanchard 2003).

Trek Aerospace (2009) "Welcome to Trek Aerospace", www.trekaero.com, accessed 2009.07.10.

Tung, Pi-Cheng, Hwang, Yean-Ren, and Wu, Ming-Chang (2002) "The development of a mobile manipulator imaging system for bridge crack inspection", Automation in Construction, 1 1 : 717-729.

Twitchell, Robert W. (2008) "Keyhole communication device for tracking and

monitoring shipping container and contents thereof, US Patent No. 7,394,361 , July 1. (2009) "Event-Driven Mobile Hazmat Monitoring", US Patent No.

7,554,442, June 30. (2010a) "Intelligent node communication using network formation messages in a mobile Ad hoc network", US Patent No. 7,733,818, June 8. (2010b) "Determining relative elevation using GPS and ranging", US Patent

No. 7,742,772, June 22. (2010c) "Using GPS and ranging to determine relative elevation of an asset",

US Patent No. 7,742,773, June 22.

Van der Auweraer, Herman, and Peeters, Bart (2003) "International Projects on

Structural Health Monitoring: An Overview", Structural Health Monitoring, 2(4): 341 -358. Vex Robotics (2009) MagBot Model 2 (www.vexrobotics.com/vex-magnetic- climbing-robot-m2.shtml, accessed 2009.06.12)

Wardhana, Kumalasari, and Hadipriono, Fabian C. (2003) "Analysis of Recent Bridge Failures in the United States", Journal of Performance of Constructed Facilities, 17(3): 144-150.

Washer, Glenn A. (2003) "Improving Bridge Inspections", Public Roads, Nov. / Dec.

2003: 51-56.

Wikipedia "Cascading Failure" (http://en.wikipedia.org/wiki/Cascading failure, accessed 2010.06.01). "Data Compression" (http://en.wikipedia.org/wiki/Data compression, accessed 2010.07.01). "Decision Theory" (http://en.wikipedia.org/wiki/Decision theory, accessed

2010.06.28). "Deepwater Horizon Oil Spill" (http://en.wikipedia.org/wiki/Deepwater_

Horizon oil spill, accessed 2010.06.30). "Deversoir" (French "Spillway";

http://fr.wikipedia.org/wiki/D%C3%A9versoir, accessed 2010.07.01). "Hyatt Regency walkway collapse" (http://en.wikipedia.org/wiki/Hyatt_

Regency_walkway_col lapse, accessed 2010.06.16). "Linear Variable Differential Transformer" (http://en.wikipedia.org/wiki/

Linear variable differential transformer, accessed 2010.06.16). "Lossy Compression"

(http.7/en.wikipedia.org/wiki/Lossy_data_compression, accessed 2010.07.01). "Multidimensional Scaling" (http://en.wikipedia.org/wiki/Multidimensional_ scaling, accessed 2010.06.27). "Photogrammetry" (http://en.wikipedia.org/wiki/Photogrammetry, accessed

2010.06.27) "San Francisco - Oakland Bay Bridge" (http://en.wikipedia.org/wiki/San_

Francisco_-_Oakland_Bay_Bridge, accessed 2010.06.01) "Statistical Process Control" (http://en.wikipedia.org/wiki/Statistical_process control, accessed 2010.06.01) "Statistical Hypothesis Testing" (http://en.wikipedia.org/wiki/Statistical_ hypothesis testing, accessed 2010.06.29)

"Statistical Process Control"

(http://en.wikipedia.org/wiki/Statistical_process_ control, accessed 2010.06.29).

Xu, Zao-Dong, and Wu, Zhishen (2007) "Simulation of the Effect of Temperature Variation on Damage Detection in a Long-span Cable-stayed Bridge",

Structural Health Monitoring, 6(3): 177- 189.

Xue, Songtao, Tang, Hesheng, and Xie, Qiang (2009) "Structural Damage Detection Using Auxiliary Particle Filtering Method", Structural Health Monitoring, 8: 101-1 12.

Yu, Zinbao, and Yu, Ziong (2009) "Time Domain Reflectometry Automatic Bridge Scour Measurement System: Principles and Potentials", Structural Health Monitoring, 8(6): 463-476.

Zeyher, Allen (2008a) "Inspectors span Hawaii for testing of bridges", Roads &

Bridges, Feb. 2008: 17.

Zeyher, Allen (2008b) "All hands on decks, piers, etc.: Tennessee mounts impressive record of inspecting bridges on time", Roads & Bridges, May 2008: 56-59.