2006-1963: SOFTWARE ARCHITECTURES FOR REMOTELY OPERABLE CIVIL ENGINEERING LABORATORIES

2006-1963: SOFTWARE ARCHITECTURES FOR REMOTELY OPERABLE CIVIL ENGINEERING LABORATORIES Prakash Kripakaran, North Carolina State University Prakash Kripakaran is a post-doctoral researcher in the applied computing and mechanics laboratory at Ecole Polytechnique Federale de Lausanne, Switzerland. His research interests lie broadly in the area of computing technologies and their applications to civil engineering. He is specifically interested in design optimization and decision support for structural engineering. He was formally a doctoral student in the Department of Civil, Construction and Environmental Engineering specializing in computer aided engineering. Abhinav Gupta, North Carolina State University Associate Professor in the Department of Civil, Construction and Environmental Engineering Vernon Matzen, North Carolina State University Alumni Distinguished Professor for Undergraduate Education, Department of Civil, Construction and Environmental Engineering; Director of the Center for Nuclear Power Plant Structures, Equipment and Piping American Society for Engineering Education, 2006 Page 11.1135.1

Software Architectures For Remotely Operable Civil Engineering Laboratories Abstract Educators have successfully adapted many classroom courses to distance education environments via the internet and are now attempting to extend this success to laboratory courses by allowing students to remotely control and observe various experiments. A key task that arises in this process of modifying the experiments in traditional laboratories for internet-enabled control and observation is the development of a secure computational framework that uses web technologies and computer networking concepts to communicate information between the computers of the laboratory and the remote user. This framework that enables internet access to the experiment must address two main issues : (1) protection for the computers that directly operate the experiment from malignant users on the internet, and (2) portability of the framework to other experiments. In this paper, we propose a framework that addresses these concerns and illustrate it for a shaketable experiment. The framework is designed to act as an intermediary between client and server applications that are developed for data acquisition and control. The key component of the framework is a proxy server. It controls access to the computers that perform data acquisition and control. A webserver that is hosted on the proxy server serves webpages related to the experiment. The webserver has user-based authentication protocols to authenticate users attempting to access the webpages. The webserver uses a combination of perl scripts and linux networking tools to setup access to the experiment for the remote user and later, disable access for the remote user when the allotted timeslot expires. Introduction In recent years, universities have witnessed a greater number of students enrolling in distance-education classes. But as the existing computing technologies are primarily designed for video-based lectures, rarely are laboratory experiments included in these classes. Laboratory experiments can be vital for students in visualizing various engineering concepts. For instance, remotely controlling a shaketable experiment may help students gain a better understanding of vibration phenomena like resonance. There have been some suggestions to use simulations in the classes to achieve the same goal. However, researchers 1 have pointed out that There will always be an important place for simulation systems, but they cannot completely substitute for experience with actual systems. For instance, simulations do not provide any insight into calibration of measuring instruments such as pressure gages and LVDT s, or into behavioral uncertainties. Several educators have proposed solutions for creating internet-enabled experiments 1,2,3,4,5. However, these solutions fail to address the key issues of security and portability. Since the experiments are accessed via the internet, the framework has to include sufficient security protocols to ensure that only authorized users are allowed to access the experiment. Moreover, the existing solutions are primarily designed to work for the specific experiment and are often difficult to extend to other laboratory experiments. For portability, it is essential that the framework consists of components with little inter-dependency. In particular, it must decouple the experiment-specific components from the components for web security, internet access, and Page 11.1135.2

web-based scheduling. To transform an experiment for remote operation, significant modifications are required. Sophisticated hardware for data acquisition and control are needed to interface measurement and control devices with a computer. Appropriate software that graphically display acquired measurements and allow manipulation of control and measurement devices must be developed. The afore-mentioned tasks pertain to the particular experiment under consideration. In addition to these tasks, novel IT architectures that create a secure web-based framework are required. The framework should also support the single controller-multiple observers concept. This capability will enable multiple users to simultaneously monitor and download measurements from the experiment as one user controls the input to the experiment. This paper focuses on the development of this generic web-based framework which together with suitably designed experiment-specific applications can transform a traditional laboratory experiment for internet-enabled control and observation. Shaketable experiment Before proceeding to describe the computational framework, we provide a brief description of the shaketable experiment for which the framework is implemented. The shaketable experiment considered in this study is primarily used as part of a laboratory course to illustrate the concepts in structural dynamics for undergraduate and graduate students. Figure 1 shows the laboratory setup of the shaketable experiment considered for illustration of the computational framework. It consists of a 12 34 one-dimensional shaketable and a 100 lb electro-magnetic shaker. The test specimen is a single or multi-story shear building having wide but thin aluminum columns and heavy steel girders. Forced vibration tests are conducted by applying a harmonic excitation to the table using a function generator. The input frequency of the excitation is increased in steps from a value that is lower than the natural frequency of the structure to one that is much higher. For each input frequency, the table is excited for a reasonable duration to ensure that the structure vibrates in steady state motion. The acceleration response is measured using accelerometers mounted on the different floors and can be viewed in an oscilloscope. This shaketable experiment was previously used by Wirgau et al. 5. They proposed a framework using National Instruments (NI) hardware and software for data acquisition and control. Programs that were developed to communicate with the NI hardware were in the form of Virtual Instruments (VI) using NI s LabVIEW development system 6. Remote access to the experiment was provided using LabVIEW s Remote Panels technology 7 and the LabVIEW webserver. The generality of this implementation is limited due to the following reasons: The implementation did not separate the experiment-specific applications, which were designed in LabVIEW, from the web technologies. A higher likelihood of intentional damage to the system by malignant users on the internet existed, since the LabVIEW host PC was directly connected to the internet. Instead of a user-based authentication protocol, the system provided security by using a list of static IP addresses. Page 11.1135.3

A scheduling facility did not exist. Such a facility would allow users to sign up for controlling or monitoring the experiment on a future date. The architectures presented in this paper are developed to address these issues. The scheduling facility is not described in this paper for brevity. In the following section, we describe the hardware setup that supports the proposed software architectures. Hardware setup of framework The hardware setup consists of two components - (1) A linux-based proxy server, (2) A local network consisting of computers that directly interact with the experimental equipment. The networking of the computers is illustrated in Figure 2 for the shaketable experiment. An IBM thinkpad with the fedora linux operating system is used as the proxy server. Figure 2 shows that the proxy server serves as the gateway to the experiment for a remote user on the internet. It has two network adapters one each for communication with the internet and the computers on the local network, respectively. While one of the network adapters is connected to the internet, the other is connected to an ethernet hub. The computers in the laboratory, which directly interact with the various experiment-related equipment, are connected to the ethernet hub. For the shaketable experiment, a National Instruments (NI) PXI that uses an NI 8176 controller and runs LabVIEW in a windows environment is used to perform data acquisition and control. It has a control board with a dedicated processor and memory. LabVIEW programs can be embedded into the board for real-time control as well as data acquisition. The accelerometers mounted on the test specimen are connected to this board through a BNC connector. Similarly, the board is also wired to the shaketable so that the generated waveforms may be communicated to the shaketable for real-time control. The PXI is connected to the ethernet hub as shown in Figure 2. Software architecture The software architecture of the computational framework essentially consists of two types of components - (1) Applications that are experiment-specific and (2) Web technologies for experiment scheduling and authentication. The interaction between these components is schematically illustrated in Figure 3. Experiment-specific applications The experiment-specific applications are those that enable communication of data to remote user as well as control of the experiment by the remote user. The proposed software architecture requires the implementation of two types of applications - (1) Server-side applications that are running on the laboratory computers and (2) Client applications that are used by the remote user. For data acquisition, the server-side applications are required to acquire the measurements from the different devices and make it available for remote users on an appropriate server. Multiple observers can use the client applications to download this data from the server. For control, the server-side application on the laboratory computer and the client application on the remote user s Page 11.1135.4

computer are required to be in continuous communication. For remotely controlling the shaketable experiment, we have used the implementation by Wirgau et al. 5. The architecture of their VIs is schematically illustrated in Figure 4. For control of the shaketable, a VI that generates the waveform as per the input amplitude and frequency is run on the host PXI. The VI on the host PXI is made available for the remote user using LabVIEW remote panels technology and the LabVIEW webserver. The remote user can control this VI through a internet browser by connecting to the LabVIEW webserver on the host PXI. In this case, the server-side application is the LabVIEW webserver and the client application is the internet browser through which the remote user accesses the corresponding VI. For data acquisition in the shaketable experiment, we have modified the LabVIEW Remote Panels-based 7 implementation by Wirgau et al. 5 to one that uses the LabVIEW datasocket technology. The proposed framework for data acquisition is given in Figure 5. The new implementation permits multiple observers to simultaneously download and monitor the measurements from the experiment. The server-side application consists of a labview VI that runs on the PXI. This VI continuously makes the measurements and writes them to a datasocket server that is running on the same computer. All users irrespective of geographical location can obtain the measurements by using a client application that subscribes to the datasocket server. The client application is a LabVIEW executable that downloads and displays the measurements in an oscilloscope. The client application provides for saving the downloaded data in a spreadsheet file for future calculations. Web architecture The server-side applications described previously for data acquisition and control operate by listening on certain network ports. The proxy server controls access to these network ports using perl-based CGI scripts within a webserver and thereby provides security to the experiment. The key actions performed by the proxy server from the moment a user logs into the system to observe or control the experiment until the time when a user disconnects from the system is illustrated using a flowchart in Figure 6. As seen from the flowchart, the following sequence of steps are involved: 1. The remote user is authenticated by the proxy server using WRAP 8, a web-based authentication mechanism. 2. The user schedules a particular timeslot for controlling or observing the experiment on a future date. 3. On the scheduled date, the user connects to the proxy server and is authenticated. 4. The user requests access to control or monitor the experiment by running a CGI script on the webserver. 5. The CGI script checks if the user is scheduled to monitor or control the experiment during the current timeslot. Page 11.1135.5

6. If the user is scheduled, the script then sets up port forwarding on the proxy server so that the user can access the appropriate server on the LabVIEW host. 7. The user runs an appropriate client application to observe or control the experiment. 8. Port forwarding that was setup in the previous step is disabled when the current timeslot expires. These steps are performed using the following components, which constitute the web architecture. User authentication North Carolina State University uses a cookie-based authentication protocol referred to as WRAP 8. This authentication protocol is used to verify the authenticity of users with access to the university s computing facilities. Students and university employees have a username and password that can be used in various computing labs around campus. Since one of the goals of this study was to make the shaketable experiment available for distance-education students, we have used the university s authentication protocol for providing internet security to the experiment. WRAP is a web-based authentication mechanism to verify the identity of a user without requiring the user to login to each individual webserver within the university domain. In this authentication mechanism, the user obtains an encrypted cookie, called the WRAP cookie, from a SSL-secured server by using his/her username and password. Whenever the user visits a WRAP-protected website, the browser sends the WRAP cookie to the website. The website verifies if the cookie is genuine and also obtains the username for that user. If the user does not already possess a WRAP cookie or possesses an invalid cookie, the user will be forwarded to the SSL-secured server that will issue a WRAP cookie to the user. WRAP cookie components like the username are available as environment variables within CGI scripts. The CGI scripts can, therefore, recognize the user making the request. The apache webserver on the proxy server is configured to use WRAP. File directories that contain the CGI scripts are protected using WRAP. Thus, the remote user is forced to obtain a WRAP cookie before attempting to run the scripts. If an authorized user is making a request, the CGI scripts use the environment variables to recognize the user. This information is later used in scheduling the user for the experiment as well as setting up remote access to the experiment for the user. Portability of framework Application of the proposed web-based framework to a new experiment involves the following tasks: Acquiring the necessary hardware that will enable the computer to communicate with the experimental equipment. Developing software applications in a client-server model for data acquisition and control. Page 11.1135.6

These tasks are experiment-specific and will be required for any experiment. But once these tasks are completed, the proposed web-based framework can merge with these software applications to make the experiment available for remote observation and control. The proposed framework needs to know only the network ports on which the client-server applications communicate. Efforts are already underway to use this framework for a cantilever beam experiment and tension test in a structures and mechanics laboratory at NC State University. Summary and conclusions Recent computing advances in web technology have given rise to new ideas in web-based experimentation for supporting engineering research and education. The National Science Foundation (NSF) is sponsoring the development of a national Network for Earthquake Engineering Simulation (NEES) that will allow large-scale laboratories with seismic testing facilities available for control and observation to geographically distributed researchers. A major task in creating remotely accessible laboratories is the development of a secure web-based framework that will support experiment-specific applications for data acquisition and control. In this study, we have proposed a computational framework with software architectures that address the key issues like security and portability. The proposed framework is illustrated for a shaketable experiment. The software architecture decouples the experiment-specific applications and the web technologies for security and experiment scheduling. The experiment-specific applications for this experiment are developed using NI s LabVIEW development system 6. These applications are designed to work using a client-server software model. The computers that operate the shaketable experiment are in a local network within the laboratory. This network is protected using a proxy server. The proxy server runs a webserver that uses NC State University s WRAP authentication protocol to provide web security. The proxy server only permits authorized users to access the experiment within their timeslot for control and observation. The main conclusions from this study are: A webserver with suitable user-authentication protocol, CGI scripts and linux networking concepts are sufficient to create a secure web-based environment to control access to laboratory experiments. The web technologies and experiment-specific applications (LabVIEW VIs in the case of shaketable experiment) are decoupled to a large extent. The dependencies between the two components are primarily with respect to the TCP/IP listening ports of the LabVIEW webserver and the datasocket server. The proxy server controls access to the experiment by enabling and disabling accessing to the TCP/IP ports on which the server-side applications operate. The computational framework is fairly generic in nature that can be easily extended to other laboratory experiments. The major task in adapting the framework to a new experiment is the development of experiment-specific applications that work using a client-server model for enabling control and data acquisition with respect to the new experiment. Page 11.1135.7

References [1] C. L. Bohus, A. Crowl, B. Aktan, and M. H. Shor, Running control engineering experiments over the internet, in Proceedings of the 13th IFAC World Congress, (San Francisco, CA), 1996. paper no. 4c-03. [2] M. L. Corradini, G. Ippoliti, T. Leo, and S. Longhi, An internet based laboratory for control education, in Proceedings of the 40th IEEE Conference on Decision and Control, (Orlando, FL), December 2001. [3] S. E. Poindexter and B. S. Heck, Using the web in your courses: What can you do? what should you do?, IEEE Control System, vol. 9, no. 1, pp. 83 92, 1999. [4] A. Gupta, M. A. Gabr, and V. C. Matzen, Alternatives in the implementation of internet-enabled laboratory, in 2004 ASEE Annual Conference and Exposition, (Salt Lake City), June 2004. [5] S. Wirgau, A. Gupta, and V. C. Matzen, Internet-enabled remote observation and control of a shake table experiment, Journal of Computing in Civil Engineering, ASCE, 2005. In press. [6] L. Wells and J. Travis, LabVIEW for Everyone. Prentice Hall, 1997. [7] NI, NI LabVIEW Remote Panels, 2005. http://sine.ni.com/nips/cds/view/p/lang/en/nid/11017. [8] NCSU, WRAP, 2005. https://webauth.ncsu.edu/wrap/. Page 11.1135.8

Function Generator Electromagnetic shaker Shaking table Shear building ẍ ẍ gr Power supply DAQ Oscilloscope Shaker arm Figure 1: Laboratory setup of shaketable experiment Proxy Server To Internet Ethernet hub PXI To shaketable Figure 2: Network setup of the computers Page 11.1135.9

Experiment specific applications Remote Computer Lab Computer Clients Server side applications Web technologies Webserver Proxy server Linux networking CGI scripts Figure 3: Software architecture PXI host PXI RT Board VI generates waveform as per user selection TCP/IP VI sends waveform to shaker Remote user controls VI using Remote Front Panels Internet Explorer Remote computer Figure 4: Wirgau et al. s framework for shaketable control Page 11.1135.10

PXI PXI host RT Board Datasocket server VI writes data to datasocket server TCP/IP VI acquires data from accelerometers Application displays data in oscilloscope Remote computer Figure 5: LabVIEW components for data acquisition To schedule Remote user connects to website During scheduled timeslot Remote user connects to website User is authorized using WRAP User is authorized using WRAP User schedules a particular timeslot User runs script to setup access User controls the shaketable via browser User runs labview application to observe experiment Controller Observer Figure 6: Flowchart showing remote user actions to access the experiment Page 11.1135.11