Garrett-research-Civil and Environmental Engineering - Carnegie Mellon University

See http://www.ce.cmu.edu/research/ais.html

Prior Activities:

2002-2003 Activity

Determining the Requirements on Product and Process Models that Support Building Commissioning

Co-PIs: O. Akin (Arch), B. Akinci (CEE)

Source of Support: NIST

The building commissioning process is a multi-phase, multi-participant process that ensures that the various interacting systems in a building are properly installed and operating at the time that a building is turned over to an owner. This process has significant benefits: improved energy efficiency; improved occupant comfort; and reduced operation and maintenance costs. There are several different product and process models for the building industry that are under development, such as the Industry Foundation Classes (IFC) being developed by the Industry Alliance for Interoperability (IAI). These models have focused primarily on the representing major tasks associated with design, construction and facility management, but not on the tasks at the interfaces of these phases. However, building commissioning has not been similarly in focus and may not be well supported by these product and process models. The purpose of this proposed project is to develop a much better understanding of the information that needs to be represented in these building product and process models to better support the many phases of the building commissioning process.

Analyzing Critical Infrastructure Dependencies: Security and Survivability Effects in the Service Sectors

Co-PIs: C. Hendrickson (CEE, PI), S. Matthews (CEE)

Source of Support: NSF CMS-0223255

Discussions about infrastructure security and survivability require system-wide comparisons and interdisciplinary approaches. Our consideration of survivability focuses on large-scale economic implications of attacks or vulnerabilities on major infrastructure sectors as defined by the Department of Commerce. We explore the critical connections between core service infrastructure sectors (e.g. telecom, electricity, and pipelines) using a total supply chain analysis model originally developed to estimate environmental and energy effects of production in the U.S.

We propose to use publicly available data from the Department of Commerce's Bureau of Economic Analysis to show the actual economic dependencies between critical infrastructures and the service economy in general. The input-output tables of the U.S. economy detail the economic purchases that result between all 500 domestic economic sectors. Specifically, these tables show how much output is generated by each of the sectors, and the amounts of product other sectors purchase from each sector. For example, the model can show not only which sectors use electricity but also how much dollar value of electricity is used to produce goods or services in that sector. These data are real and tangible representations of how dependent service sector businesses are to other businesses within the infrastructure sector and the economy in general.

Micro-NMR device for Detecting Chlorides in Concrete

Co-PIs: P. Sides (CHE), G. Fedder (ECE), M. Patton, I. Oppenheim and I. Lowe (PITT)

Source of Support: NSF CMS-9980759

Monitoring chloride concentration and transport in concrete structures susceptible to corrosion of embedded steel reinforcement is a challenge as difficult as it is important. An embedded sensor based on nuclear magnetic resonance (NMR) would be a good solution to the problem because it would make a non-destructive atom-specific measurement of the presence and concentration of chloride. The important question is the scale of the device required to detect the chloride. Laboratory experiments to detect chloride in a cement matrix using pulse-NMR were conducted to assess the potential of this application; they provided a basis for projecting the scale of a device that would have a good chance of success. The coils were cm-scale and the magnetic field was 2.35 T. NMR signals were obtained from both aqueous chloride solution and samples of both regular and white portland cement. The experiments demonstrated that the signal-to-noise ratio (SNR) for a cm-scale cement sample volume is so small, even after averaging, that sample volumes much lower than that are unlikely to produce measurable signals at fields of 1 T or below. Thus, the potential for realizing an embedded NMR-based sensor including the magnet is low. Parametric studies identify feasible alternative coil diameters and magnetic field strengths for detecting chloride ion concentrations in hardened concrete. Another aspect of this project is to develop a MEMS chip capable of force detection NMR (FDNMR). The project team is in the process of developing this MEMS chip using surface micromachining.

Knowledge Discovery and Datamining for Civil Infrastructure Contexts

Co-PIs: Rebecca Buchheit (CEE), Sue McNeil (UIC) and Christos Faloutsos (SCS)

Source of Support: NSF CMS- 9987871

In the first year, we have devoted our efforts to: (1) identifying the types of data quality errors that frequently occur in civil infrastructure monitoring data; (2) developing a data quality assessment procedure to identify these errors; (3) testing the assessment procedure on case studies; (4) systematically testing the sensitivity of the data quality assessment procedure; and (5) developing a software architecture for the data quality assessment procedure.

We have identified eight types of errors that occur frequently in civil infrastructure monitoring data. These errors can be divided into two groups: systematic errors and individual record level errors. Calibration errors, threshold errors, missing data, and extra data are common systematic errors. Missing or incorrect values within records, two records that are combined into a single record, a single record that is split into two records, and duplicate records are common errors that occur at the level of individual records. We have developed a two-level data quality assessment procedure to identify these errors. In the first level of the procedure, several different data quality assessment methods are used in a voting scheme to identify concentrations of anomalies in aggregate data. In the second level, differences between anomalies and normal data at the individual data level are identified; combined with domain knowledge, these differences can be used to identify different types of errors, such as missing data and calibration errors. Then, the data can be cleansed effectively.

So far, we have tested the data quality assessment procedure using two case studies. The first case study uses weigh-in-motion data from the Minnesota Department of Transportation; the data consists of the weight, length, and speed characteristics of trucks traveling on Interstate-94 westbound. The second case study uses Heating, Ventilation, and Air-Conditioning (HVAC) data from the Intelligent Workplace building, which is located on the Carnegie Mellon University campus. We have also developed a test bench to systematically explore the sensitivity of the data quality assessment algorithms. The test bench introduces a known error into a clean, artificial data set and then evaluates how well each assessment method identifies the error. In order to automate the data quality assessment procedure, we have begun to develop a prototype software architecture. Currently, the prototype implements a constraint-based evaluation system; the system allows the user to specify physical constraints for the data (i.e., the gross weight of a vehicle must be less than 80 kips), as well as constraints on the distributions of attribute values (i.e., a normal distribution with a specific mean and variance).

2000-2001 Activity

Micro-NMR device for Detecting Chlorides in Concrete

Co-PIs: P. Sides (CHE), G. Fedder (ECE), M. Patton, I. Oppenheim and I. Lowe (PITT)

CEE Students: Andrew Yun (PhD); Source of Support: NSF CMS-9980759

The ultimate objective of this research is to develop an inexpensive passively powered NMR system that can be distributed throughout a volume of concrete during casting and monitor the concentration of chloride ions in the surrounding concrete. Late this past fall, Andrew Yun was able to detect small (0.1% by weight of portland cement) concentrations of chlorides in gray portland cement using a 2.35T NMR system in Irving Lowe's laboratory. We have not seen any literature describing NMR signals for chlorides below 9T. We also were able to show that some chloride that might be detected by existing means may actually be bound and unavailable for corrosive activity.

MEMS-based Sensors for Infrastructure Health Monitoring Applications

Co-PIs: Mark Patton;

Source of Support: NSF, Pennsylvania Infrastructure Technology Alliance (PITA)

During this past year, we started projects with High Concrete and Microstrain to explore the use of microdisplacement transducers, RFID tags, and 916 MHz wireless microtransmitters for monitoring the health of precast concrete T-beams and wall panels as they proceed through the casting process and placement in the curing yard. Mark Patton has already been able to illustrate the value of such in situ health monitoring to High Concrete; in late fall, he was able to embed strain sensors in the precast T-beams and detect when cracks occurred during the release of the pretensioning cables from the forms. This was an unexpected result according to High Concrete.

Speech-Controlled Wearable Computers for Supporting Field Workers

Co-PIs: Markus Klausner; CEE Students: Jirapon Sunkpho, Christian Buergy, Jan Reinhardt

Sources of Support: Bosch (Christian), PITA (Jirapon Sunkpho), Hochtief Construction (Jan Reinhardt)

This area of research has focussed on the development of effective speech-controlled applications supporting field workers, such as bridge inspectors, automotive inspection and manufacturing workers, and construction project managers. Jirapon Sunkpho is developing an object-oriented framework that will allow application developers to more quickly and easily prototype Java-based speech-controlled applications for inspection-oriented applications. Christian Buergy has worked this past year to develop several very successful speech-controlled automotive inspection applications, in German, that have been tested at Bosch automobile inspection facility in Germany. We have submitted a paper on this project to the SAE 2001 World Congress (see section 4.2).

Knowledge Discovery and Datamining for Civil Infrastructure Contexts

Co-PIs: Sue McNeil and Christos Faloutsos (SCS)

CEE Students: Becky Buchheit (PhD); Source of Support: NSF GRT

The primary objective of this research project is to develop a framework for systematically applying the knowledge discovery and datamining (KDD) process to civil infrastructure data analysis needs. Based on several civil infrastructure-based case studies that have been performed to date and reported on in the literature, the most difficult and important steps in this process are preparing the data and evaluating its quality. Becky Buchheit is developing a framework to support these two steps in the datamining process as her PhD dissertation work.

Advanced Sensors for PeopleMover Guideway Construction and Operation

Co-PIs: Burcu Akinci, Chris Paredis, Scott Thayer

CEE Students: Richard Hallon (MS), DeWitt Latimer (MS); Source of Support: PITA, (year 3&4), Adtranz NA

For his MS thesis, Richard Hallon explored a variety of sensors that could be used to detect ice formation  on roadway surfaces. He did a set of experiments with a IR sensor being developed by a Swedish  company for the Adtranz PeopleMover Guideway context. Late this year, we initiated a project with Adtranz to investigate the applicability of laser scanning technologies for surface assessment of  PeopleMover guideways; DeWitt Latimer will be working on this project for his MS Thesis. DeWitt is being co-advised by Prof. Burcu Akinci.

1999-2000 Activity

Micro-NMR device for Detecting Chlorides in Concrete

Co-PIs: P. Sides (CHE), G. Fedder (ECE), M. Patton, I. Oppenheim and I. Lowe (PITT)

CEE Students: Andrew Yun (PhD)

The ultimate objective of this research is to develop an inexpensive passively powered NMR system, with a magnet about the size of a small egg and a Micro-NMR chip inside it, that can be distributed throughout a volume of concrete during casting, that can monitor the concentration of chloride ions in the surrounding concrete, and that can store, process, filter and communicate the sensed data to the surface when interrogated. Detection of chloride penetration into concrete is an important function for maintenance of the nation’s surface transportation infrastructure. $50 - $200 million are spent per annum to repair or replace concrete bridge decks damaged by corrosion due to chloride attack. The project, currently supported by the National Science Foundation's XYZ on a Chip Program, will be focussed on the proof of the principle of miniaturization and on parallel application-oriented design studies necessary to establish feasibility of the idea.

MEMS-based Smart Aggregate
Co-PIs: K. Gabriel (ECE) and M. Patton

CEE Students: Douglas Sackin (MS), Richard Hallon (MS)

This research project is exploring the possibility of using MEMS-based sensors and wireless communication technologies as a basis for distributed monitoring of concrete structures, primarily bridges and pavements. The approach and many of the technologies can be applied to other types of infrastructure, but concrete provides both a major market and a platform to explore in situ needs. Prototype development is centered on the integration of commercially available technologies. During this past year, we conducted numerous experiments with existing devices: MEMS accelerometers and pressure sensors, IC temperature sensors, RFID tags, and 916 MHz microtransmitters. In several different experiments we embedded these different sensors in concrete test specimens and observed their performance as the concrete cured. In the case of the microtransmitter, we observed the change in the transmission distance as the concrete was cured. For next year, we intend to focus this research activity on a potential application with Adtranz, funded by PITA and Adtranz, which is related to monitoring the concrete they use in their guideways.

Speech-Controlled Wearable Computers for Supporting Field Workers

Co-PIs: Markus Klausner and Asim Smailagic (ICES)

CEE Students: Jirapon Sunkpho, Christian Buergy

This area of research has focussed on the development of effective speech-controlled applications supporting field workers, such as bridge inspectors. Jirapon Sunkpho has been working on a bridge inspection application for several years. During this past year, he began work on his dissertation topic: to develop a framework that will allow application developers to more quickly and easily prototype speech-controlled applications for inspection-oriented applications. Christian Buergy has worked this past year to develop a very successful speech-controlled application, in German, that will be tested at Bosch facility in Germany in early February. This project is leading to additional research activity related to exploring technologies that will reduce the size of the hardware, with support from Bosch labs in Palo Alto, and exploring effective audio-centric application interfaces.

Knowledge Discovery and Datamining for Civil Infrastructure Contexts

Co-PIs: Sue McNeil and Christos Faloutsos (SCS)

CEE Students: Becky Buchheit (PhD)

The primary objective of this research project, which is Rebecca Buchheit's PhD dissertation topic, is to develop a framework for systematically applying the knowledge discovery and datamining (KDD) process to civil infrastructure data analysis needs. The framework will act as both a checklist and a set of guidelines that will interactively give suggestions to help researchers choose the most promising data mining techniques, prepare their data, and consistently evaluate the results of the data mining techniques. To date, Becky Buchheit has conducted several KDD case studies in civil engineering-oriented domains: (1) an earth moving operation in northeastern Pennsylvania; and (2) the Intelligent Workplace and the data that was collected for a year concerning power usage, HVAC system operation, lighting, etc.

1998-1999 Activity

Microelectromechanical Systems (MEMS) Applications in Civil Infrastructure
(Doug Sackin, Jim Garrett and Ken Gabriel)

Infrastructure plays an important role in society and yet we are often limited in the information we use to make important decisions regarding repair and replacement. Over time, structures deteriorate as a result of use and environmental factors. Construction errors may also cause the initial properties of a structure to be unpredictably different from those specified by the design.

Inspection and testing methods are available for evaluating structures but they are typically employed infrequently due to the expense and difficulty of their use. Many of these methods permit only surface measurements that may not be representative of the bulk of a structure or else may require some amount of destruction, such as the collection of core samples. Furthermore, most existing structures are not designed and built for inspectability. If structures cannot be inspected adequately, repairs may be executed too soon, leading to unnecessary expense, or too late, leading to devastating and costly failure.

To address these problems, we are developing microelectromechanical systems (MEMS) for embedded infrastructure monitoring. MEMS devices are micromechanical structures that can embody both mechanical and electrical functions. The devices can be manufactured using the same materials and processes as microelectronic devices such as microprocessors and memory chips. These manufacturing methods permit the devices to be small enough and inexpensive enough to distribute throughout a structure in numbers from hundreds to millions. The common manufacturing process also permits the integration of mechanical actuators with electronics for control, processing, and communication. This combination introduces the ability to directly sense and interact with the structural environment.

There are several advantages of a MEMS-based monitoring system over other methods of condition assessment and monitoring. MEMS devices can be embedded, bringing their readings closer to the true in situ properties of the structure and not estimates based on external visual inspections or expensive nondestructive tests. MEMS technology has the potential to measure the properties of interest directly. Batch processes are used to fabricate MEMS devices, introducing high quantity and low cost and ensuring good coverage of a structure. Finally, MEMS devices are robust and bring improved ease of use. Their small, encapsulated nature makes them difficult to damage while wireless technologies make them easy to embed and interrogate.

A MEMS-based system for distributed infrastructure monitoring requires integratable technologies for sensing, power, wireless communication, device location and orientation, computation, and storage. In addition to the individual embedded devices, procedures must be developed for design, distribution, interrogation, and data analysis for condition assessment.

This research project examines each of these issues as demonstrated by the implementation of a MEMS-based system for distributed monitoring of concrete structures, primarily bridges and pavements. The approach and many of the technologies can be applied to other types of infrastructure, but concrete provides both a major market and a platform to explore in situ needs. Prototype development is centered around the integration of commercially available technologies. Once a platform for the core capabilities, power and communication, has been developed the devices can be adapted to other domains by customizing the sensing mechanisms. In this prototype, inductive coupling provides the platform power and communication.

We refer to the devices as Smart Aggregate and envision them to be the size of coarse concrete aggregates (approximately 2.5 cm in diameter). They are distributed into the structure during construction for cure and, later, condition monitoring. Future work will address the need for retrofitting the devices into existing structures.

Each device has a suite of mechanisms for assessing the condition of the concrete. Some of these sensors, such as temperature and acceleration, are existing microelectronic and MEMS components that can be directly integrated. Other sensors are new MEMS devices developed for gauging other properties in the concrete, such as those related to strength. This combination of mechanisms supports the application of several existing nondestructive evaluation methods in addition to the capabilities of each individual sensor.

MEMS-based devices provide us with the ability to inexpensively monitor structures from the inside out. With this in situ information we can better manage our vital infrastructure and research better materials and practices.

Wearable Computers for Bridge Inspection
(Jirapon Sunkpho, Jim Garrett, Asim Smailagic, Dan Siewiorek)

Wearable Computers for Bridge Inspection Presentation (Jan 1998)

The objective of this project is to assess the needs of bridge inspectors in the field, to develop a set of prototype information access and retrieval devices that require nearly hands-free operation, and to field test these devices with actual inspectors to demonstrate the benefits and costs of this approach. The project started in the fall of 1997, when we interviewed several inspectors from the Baker Corporation and PENNDOT District 11. Based on these interviews, we then defined the design context for the Spring 1998 course on wearable computer design. The first prototype developed incorporated speech recognition software, a hand-held LCD display, support for digital photography and sketching, and wireless communication capabilities. We showed the prototype (based on CMU hardware) to the PENNDOT inspectors and engineers and got their feedback. We spent the Fall 1998 exploring new hardware platforms and means for improving the performance of the speech recognition system based on the structure of the inspection context. The potential impact of this work is significant as almost all State DOTs are still looking for the good systems to use in supporting their inspection field operations. PENNDOT is going to fund a significant field test in Pittsburgh and the Lehigh Valley for the summer of 1999.

EIOLCA.NET
(C. Hendrickson, Jim Garrett, A. Horvath, S. Mathews)

In this project, the objective has been to develop a web-based version of a tool for performing life-cycle assessments (LCAs). The Economic Input-Output Life-Cycle Assessment (EIOLCA) tool has been the subject of research and development within the Green Design Initiative for the last several years. This past year, the NSF and the EPA co-funded a project for which one of the objectives was to explore bringing this tool to the web. We have delivered a web-based version of this software (see URL eiolca.net). To make this system able to respond quickly to web-based requests for EIOLCA analysis results, we designed the software to compute and cache the results so that as the system is used it will become faster.

See EIOLCA.NET

Environmental Standards Broker
(Jim Garrett, Vei-Chung Liang and Steve Fenves)

Broker Webpage

We have been developing a prototype WWW-based broker for providing access and support for using environmental regulations. The broker supports several types of searches: full-text searches; classifier-based searches; and browsing. Full-text searching is useful if the words provided in the user query appear in the text, but it does not aid in finding implied concepts. To aid in the latter, the broker also incorporates a classification system. This year, we completely redesigned and reimplemented the prototype broker as a Java applet and Java server application. This redesigned broker freed us from our dependence on the very expensive Inquery software that we were using to do the classifier and full-text searches in the previous versions of the broker. Based on this new version of the broker, we have developed a pilot application for the Pennsylvania Department of Environmental Protection (PADEP) that supports usage of their Air Quality Policy documents. This pilot is currently in field test; PADEP air quality engineers are evaluating the functionality and performance of the system. Based on this evaluation, we will develop a contract with the PADEP and the State to further develop and test additional broker functionalities. These functionalities will require additional research and development regarding information retrieval and the opportunities to apply AI and machine learning techniques.

Graphical Interactive Website for Teaching/Performing Planar Truss Analysis
(Richard Quinn and Jim Garrett)

We designed and developed a Java applet that interacts with students via the Web, allows them to graphically specify a truss analysis problem, generates a truss analysis file and ships it off to an FEM analysis server (using FElt). The results of this analysis are then graphically displayed to the students in their web browser in several different ways. This software was used to introduce the concept of planar truss behavior and computer-based structural analysis to the students in the 12-100: Intro to CEE course in the Fall 1997, 1998 and 1999 semesters.

Engineering Design Standard Representations and Processing
(Han Kiliccote, Bongjin Choi and Jim Garrett)

Because of the complex dynamic nature of design standards, combined with the need to incorporate standards into computer-aided design systems, a major goal of this research has been to deliver vendor-independent software support for the usage of design standards in engineering design software. The vision for delivering standards usage support that has been guiding my research has formal, computable models of standards being served over the Internet from servers operated by standards promulgating organizations. Design and evaluation software would then communicate with these servers in order to provide standards usage support. Towards this objective, we have created a prototype agent-based framework on which to provide these standard usage support services. We call this framework the Standards Processing Framework (SPF). All agents within the SPF communicate via the same language, known as the SPF Communication Language (SPF-CL). Using SPF-CL, it is possible to define a model of a standard, define a design context to check for compliance, and query the results of this compliance check. The SPF thus provides a standards compliance checking service with which various design software packages can interact to provide standards checking services to their users. Our plans our to apply and validate this approach in several contexts: the SEED project; the USACERL Modular Design System; and an American Institute of Steel Construction (AISC) Steel Specification Processing System.

Machine Learning for Support of Early Collaborative Design
(Nenad Ivezic and Jim Garrett)

The objective of this research project was to develop an approach for bringing the knowledge embodied in detailed simulation tools into the early stages of design. Ivezic and I developed an approach where a neural network is used to create a model of the relationships between the form and behavior of a design, using examples generated within the design space and simulated using the detailed simulation tools. Once created, this model of behavior can be used during the early stages of design to provide feedback to the designer about the likely behavior that would result from the design decisions made. Conversely, the same model can be used to identify decisions about the form of a design that would need to be made to yield a desired behavior. Ivezic developed a prototype system that illustrated this approach.