Page 121

D

Illustrative Science Base and Scope for Remote Technology for Decontamination and Decommissioning of DOE Nuclear Facilities

Background

The decommissioning and dismantlement of nuclear facilities entails a wide range of physical tasks (e.g., inspection, cutting, handling, packaging) of a very diverse set of structures (e.g., piping, valves, pumps, tanks, wire conduits, building structures, concrete). Successful D&D activity is possible today using existing technology based primarily on manual operations, which entails high cost and long timelines. Numerous attempts have been made to use remote systems technology (robotics) to reduce hazards to workers and to reduce the cost of the total life cycle of the operation. Barriers to the deployment of an advanced technology arise from a history of taking a low-risk approach, the deployment of industrial robotics technology poorly suited to the functional spectrum associated with D&D, or the development of on-off devices (with a nominal record of performance).

All of this disparate activity suggests the absence of a national strategy (based on adequate resources) for a science base that would aggressively attack the technology needs with a balanced approach where timely investment can be made to address weaknesses and opportunities to most rapidly deploy a cost-effective technology. The existing project-oriented approach results in special-purpose systems developed under contract primarily by industry to meet a local site need independent of long-term research activity at universities and the DOE's own research laboratories. Further, this special-purpose technology depends on a special class of maintenance technicians provided by the system supplier



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 121
Page 121 D Illustrative Science Base and Scope for Remote Technology for Decontamination and Decommissioning of DOE Nuclear Facilities Background The decommissioning and dismantlement of nuclear facilities entails a wide range of physical tasks (e.g., inspection, cutting, handling, packaging) of a very diverse set of structures (e.g., piping, valves, pumps, tanks, wire conduits, building structures, concrete). Successful D&D activity is possible today using existing technology based primarily on manual operations, which entails high cost and long timelines. Numerous attempts have been made to use remote systems technology (robotics) to reduce hazards to workers and to reduce the cost of the total life cycle of the operation. Barriers to the deployment of an advanced technology arise from a history of taking a low-risk approach, the deployment of industrial robotics technology poorly suited to the functional spectrum associated with D&D, or the development of on-off devices (with a nominal record of performance). All of this disparate activity suggests the absence of a national strategy (based on adequate resources) for a science base that would aggressively attack the technology needs with a balanced approach where timely investment can be made to address weaknesses and opportunities to most rapidly deploy a cost-effective technology. The existing project-oriented approach results in special-purpose systems developed under contract primarily by industry to meet a local site need independent of long-term research activity at universities and the DOE's own research laboratories. Further, this special-purpose technology depends on a special class of maintenance technicians provided by the system supplier

OCR for page 121
Page 122therefore severely limiting operational flexibility at any given site. By contrast, with a national strategy, generalized technology could be deployed to not only cover all relevant applications but could do so at low risk and dramatically reduced costs. In fact, it is suggested that deployment costs can be reduced by more than 50 percent, and using sophisticated training and operational software, timelines could be substantially reduced, and overall logistics functions (maintenance) also could be dealt with as a cost-cutting opportunity. How Can This Be Done? First, central computer controllers now have an open architecture and their cost is coming down rapidly while their performance is increasing. The expected controller cost with software for the open architecture remote systems is expected to be based on continuously advancing personal computers and to cost less than $10,000. Second, even though a very broad range of remote systems will be required (handling, manipulators, rovers, etc., all at various scales), they can be built with an open architecture based on a collection of about ten standard actuator modules with an average cost less than $5000 in reasonable quantities (100+). These actuators are 80 percent of the remote system. They contain all the wiring, motors, brakes, local electronics, sensors, interfaces, etc. Because of their standardized interfaces, they can be rapidly assembled on demand to create a wide population of remote systems to meet any specific application. Third, because of the standard interfaces, the actuators can be removed and rapidly replaced and a minimal set of spares kept on hand to maintain the system. Also, expensive, highly trained service personnel become a thing of the past (just as has occurred for computer systems). Hence, life-cycle costs come down a great deal. These standardized interfaces would also apply to sensor modules (used for site characterization), process tools (for size reduction), and to communication subsystems (power umbilicals, wireless data transfer, etc.). Fourth, because the remote system is made up of modules having standardized interfaces, tech mods (better motors, sensors, brakes, embedded electronics, controllers, etc.) can be inserted to improve the subsystem without disturbing the standards and done so at very low cost. Hence, the up-front risk to the designer, project decision maker, and site contractor goes way down. Once the standards are accepted,

OCR for page 121
Page 123in-depth qualification tests of these subsystem modules becomes possible, certifying their performance metrics for a wide range of operational conditions. This is equivalent to the lessons learned from the market for personal computers: performance goes up while costs go down. Fifth, it is now possible to establish a universal operating software to operate all remote systems that can be built from the standardized subsystems. This is a lesson learned from Microsoft. We need to apply it to mechanical systems as well as for computer-based systems. The robotics research community has in place a broad body of analytics to support decision making required to operate complex systems (from 1 to 20 degrees of freedom), to expand their performance envelope, and to allow the judgment of the human operator to efficiently manage this complex system through a spectrum of organic interfaces. Economics and Program Planning. D&D represents an estimated life-cycle cost to DOE of $33 billion, of which only 12 percent will be performed before FY 2007. Hence, there is a long timeline in which to properly develop cost-effective technology. A good long-term program for development of remote systems has been developed by the DOE community and others in the form of the Robotics and Intelligent Machines (RIM) roadmap. This program lays out a strategic timeline for overall performance goals and targets for a demanding list of D&D applications (as well as mixed waste, nuclear materials, etc.). One of these performance goals is to reduce worker exposure to radiological hazards by 90 percent while increasing productivity by 300 percent for such applications as the Hanford building 324 hot cells, the Rocky Flats glove box size reduction, etc. This plan recognizes the need for advances in certain component technologies (planning in unstructured environments; characterization sensor suites, data analysis, management, and visualization; specialized manipulators and handling equipment; mobile platforms and their navigation; configuration management; man-machine interface and operator training; tetherless operation; etc.). However, the RIM plan does not offer a science strategy to respond to these valid program goals. Remote Systems Science Activity To Support Long-term Development for D&D. The following list of topics provides a reasonable description of the technologies and the required science to support advanced remote system development for D&D. Each is described in terms of the most

OCR for page 121
Page 124promising development and, in many cases, represents a major transition from present research in the topic. Electronics Hardening. Semiconductors continue to undergo intense development of smaller geometries, higher doping, reduced operating voltages, ultra-thin active regions, reduced dielectric strengths, and developments in device structures (hetero-junction bipolar transistors, high-electron mobility transistors, quantum well, super lattice, v-groove, etc.), which offer many improvements for radiation tolerance. In-depth understanding of the device physics must be developed to create models as the basis for the design of new technologies for radiation tolerance. Empirically, we can confirm that newer electronic components are more tolerant to terrestrial radiation (total dose) environments but at the expense of a decreased Single Event Phenomena (SEP) capability (important in low-radiation fields). As geometries become smaller, multiple errors as a result of a single ion strike and new potentially destructive mechanisms such as a rupture of the oxide layer may cause operational problems. A challenge exists to develop a science base that would predict when a device will no longer tolerate the multiple effects from radiation, especially as the complexity of these devices increases. Recent development of MESFET GaAs technology and bipolar technologies suggest that total dose radiation tolerance will be greatly improved in the future. For the system designer, it becomes essential to have explicit knowledge of the radiation tolerance of any given device, now realizable only by continuous testing and evaluation. Hence, a science of tolerance measurement of a broad range of devices, the acquisition of lessons learned, and the reduction of these data as guidelines to the electronics designer now becomes highly desirable. Sensors for Characterization and for System Operation. Here, the principal objective is to create a new level of science to accelerate the technology deployment for the mobile and remote mapping of large indoor and outdoor areas consisting of pipe and tank-like structures. This includes the processing of sensor data to identify objects within the acquired scenes (including radiation levels and sources) as well as aid in the visualization of the scene by the operator. These technologies include data registration, multi-sensor data fusion, scene building, object and scene modeling, object detection and recognition, and sensor characterization (see Chapter 4). Specific research tasks might be listed as follows:

OCR for page 121
Page 125 Task D&D Application 1. Next Best View in 3D Object Modeling Uses only a few views, without prior knowledge, permits object modeling. Use as aid to robot in identifying unknown objects in D&D structure to be dismantled. 2. Registration of Multi-Sensory Modalities Automatically register multi-modal 2D projective data sets of a scene to a global/3D coordinate system (including range and color data). Useful for mapping robot's work environment and providing collision data for remote operation of multi-degree of freedom mobile robot and manipulator. 3. Multiple View, Multiple Wavelength, Thermal and Radiation Imaging for 3D Object Characterization Simulated data capture for thermal and radiation sources, construct time varying scenes as the source levels vary. Data can be mapped as a texture onto 3D object scenes. Multiple wavelength infrared and radiation cameras permit true temperature and radiation data capture without tedious calibration. 4. Multi-modal Measurement for Image Collage Register multiple range and multi-modal images to a common reference frame, integrate into a unified, textured, 3D model for display. Permits D&D scene construction to guide the robot for inspection and sequential structure dismantlement. 5. Data Reduction of 3D Meshes for Multi-Resolution Analysis Using Wavelet Transform Use results of multi-resolution analysis to guide a mesh reduction strategy, use quincunx wavelet transforms. D&D scenes can produce data magnitudes that can be over-whelming. This approach retains only sufficient data to enable practical visualization and processing. 6 Object Modeling in Multiple Object Scenes Using Deformable Simplex Meshes Automatically models 3D point clouds that might comprise multiple objects using deformable simplex meshes, shrinking the mesh according to physical constraints, then refining the mesh to identify multiple objects allows determining the position and orientation of complex pipe structures, valves, even fine segmentation of small objects as can be expected in the D&D work environment. 7. Volumetric Primitive Object Presentation of Range Images for Object Recognition Recover geometric models for each part of each object from range image data, reduce noise and background clutter, identifying objects with finite number of primitive models. Enhance robot operations on repetitive (similar objects) removals from the D&D environment.

OCR for page 121
Page 126 Mobile Platform Navigation and Operation. Position estimation deals with the ability of a mobile robot to estimate its position relative to the environment (both known and unknown). GPS provides typical accuracy of 10 meters which has little value for D&D operations and is altogether not usable indoors. For this reason, scientific development is required on systems employing external references (such as beacons and artificial landmarks) and on dead-reckoning systems. To date, all existing navigation systems have distinct shortcomings for use on D&D. Obstacle avoidance is a serious problem. The need is to detect obstacles in time to avoid collision and to circumnavigate the obstacle. Research is warranted in two main areas: computer vision-based systems (i.e., those that employ cameras) and those that employ range sensors. The focus in computer vision performance is suggested to be on advanced computer software. For range sensing, continued development of LIDAR and FLASH LIDAR sensors (using range snapshots providing rich information for all objects within sight of the sensor) seems promising. Dramatic improvements in the kinematic design of mobile robots will be necessary before truly versatile performance can be expected. In unstructured environments (as found in D&D), it is quite difficult to be fully functional on wheels only (wall climbing, internal pipe operation, traversing rubble, etc.). Alternative methods of propulsion are usually less efficient and incur penalties in weight, complexity, control, and accuracy. For example, legged robots are slow, fragile, and cumbersome in operation. Suggested research would include a reconfigurable system that could be rapidly assembled from standard modules, enabling the utilization of several distinct modes of propulsion or even combinations. Virtual Reality for Enhanced Man-Machine Interface and Training. In the initial planning and characterization phases of D&D work, workers must enter an area of high radiation and contamination that is also congested with left in place equipment and materials for which removal inevitably involves physical stress (fatigue) and the potential of personal injury. Virtual reality systems combined with mobile robot platforms (including advanced navigation technology) could allow workers to perform essential survey and decision making functions from a remote location, thus enhancing their safety and productivity. Advances in the state of the art as now used in deep sea exploration (navigation, scene analysis, specialized sensors) should be pursued to improve overall system performance by means of force feedback (of physical operations), remote vision (for registration, object locations, radiation sources, etc.), collision avoidance (because of the lack of definition of the work area), and radiation-resistant system technology (especially the sensors them

OCR for page 121
Page 127selves). The required science involves an enduring issue of a balanced technology to enhance the relationship between the operator and the machine. This is partially due to the increasing complexity (options and performance) represented by the machine and the desire to perform more complex tasks. The required science must deal with the following questions. What are the most efficient channels of communication (visual, sound, kinesthetic), which channels can easily be overloaded (too much information flow not easily interpreted by the operator), which can be used to reduce the potential for operator fatigue (both physically and mentally), enhance training, provide in-situ skilling, and overall can this technology reduce costs specifically for the characterization and physical tasks associated with D&D. Intelligent Actuators. We all recognize the pervasive influence of the computer chip in modernizing our information technology. The equivalent in intelligent machines is the actuator. We need to embed the same level of excellence in component and system technologies in the actuator as we now do in the component technologies for computers. Only then will we dramatically improve performance while reducing costs. Only then can we consider building fully integrated machines on demand (as we now do for computers). Here, we recommend a concentration in all the sciences to modernize actuator technology so that it stands as an equal partner to the computer chip in order to create a balanced technology for a broad population of remote systems for D&D. The following listing of scientific concentrations is recommended to dramatically impact the level of technology for intelligent actuators. 1. Advanced Component Technology —One of the most troublesome components in an actuator is the gear train; it needs to be modernized to be lighter, stiffer, easily manufactured, etc. Similar efforts are required for brakes, bearings, prime movers, magnets, wiring insulation, etc. Also, standardized quick-change interfaces of high precision need to be developed of high rigidity and lowered cost. 2. Micro-Sensors —Unfortunately, almost all existing actuators contain only one sensor, making it impossible to make the module intelligent. Here, we recommend the embedding of 10 (or more) sensors in every actuator (temperature, torque, current, voltage, position, etc.) to generate a full awareness of the actual condition of the actuator. These sensors should benefit from

OCR for page 121
Page 128 advanced MEMS technology and represent the best production experience now used to make sensors for automobiles (i.e., very small, rugged, low cost, etc.). 3. Actuator Design —Literally hundreds of parameters will be involved in the design of these actuators. A process must be put in place to interactively control these parameters so that advanced structures will result to meet the most demanding application requirements (extreme low weight, high power density, good disturbance rejection, etc.). This includes the design of the prime mover, choice of materials, method of fabrication, etc. 4. Operational Software —Present actuator technology depends primarily on an outdated concept of feed-back control. What is needed are criteria-based decision making systems to maximize performance, to make it possible to provide condition-based maintenance for maximum system availability, and to provide a fault tolerant architecture to reallocate resources inside the actuator to continue system operation even in the presence of a fault. 5. Maximum Performance —Most actuators are conservatively operated to prevent saturation. Here, we wish to maximize performance just as we do for the engines of our fighter aircraft. It is recommended to test each actuator as-built, document all parameters, useful operational criteria, their norms, combinations of criteria to meet specific performance objectives, conflict resolution among performance objectives, etc. All of these issues should be prioritized by the operator in the field, archived for lessons learned, and to advise the operator relative to the performance reserve available from a given actuator. 6. Layered Control —Virtually all present actuators operate at only one scale. The biological system has been shown to operate at three to four distinct levels of motion. It is recommended to create a similar set of scales in the full architecture of these actuators, making it possible to maintain a much higher level of accuracy even under external disturbances. Special use of MEMS technology will make layered control possible. 7. Ultra-Light —Frequently, applications occur where actuators of considerable torque capacity must be extremely light. It is now possible to aggressively reduce the weight of these actuators,

OCR for page 121
Page 129 while still making them shock resistant as well as extremely compact. The best selection of materials, optimum selection of prime mover technology, and highly structured design rules and procedures can address these issues to meet this objective. 8. Fault Tolerance —Normally, fault tolerance is achieved by duplication of actuators—one dormant while the other operates—a complete disaster for efficiency, weight, and cost. Here, it is recommended that a dual system fully integrated be developed where all components are used at all times for maximum performance with no single point failures. Should a partial failure occur, the system would be reconfigured based on criteria to minimize the affect of the failure and still achieve nearly optimum performance. This capacity is essential where access to the remote system is difficult and where endurance and long-term availability are paramount. 9. High Efficiency —Certain applications demand the use of the absolute minimum of power, primarily to minimize the weight and cost of portable power packs. Generally, this requirement implies very little heat generation by using criteria to reduce spurious responses, peak torques, hysteresis in magnetic materials, etc. An intelligent actuator based on 10 sensors is the ideal technological environment in which to treat extraordinary requirements of this class. Universal Operating Software for D&D Systems. The suggestion here is to develop the science base for universal operational software for mechanical systems that are increasingly becoming more intelligent in response to a need for greater performance, flexible response to changing requirements, fault tolerance for continued operation under a fault, and condition-based maintenance to enable repair by rapid module replacement by a nominally trained technician—all to be achieved at reduced costs. The crucial reality is that a gigaflop controller technology exists today as a $5000 commodity to provide immense decision making resources in real time to allocate resources within increasingly complex mechanical systems. By the year 2010, it is predicted to be 50 gigaflops. This means that the simple analog concept for “control” of our mechanical systems must now give way to a sophisticated criteria-based “management” of excess resources (options) to maximize the system's response to human intervention and to create open architecture systems that can be rapidly reconfigured to respond to a fault, to integrate a new tech

OCR for page 121
Page 130nology, and to maximize performance (at reduced cost). Elements of this approach have been pursued for our fly-by-wire aircraft, automobiles, household appliances, etc. In order that we create the most flexible, responsive, and cost-effective technology for D&D, it is essential that these unique software requirements be met to augment the science of the software field itself. Increasingly, our most advanced mechanical systems are being controlled by a predictive model reference based on as-built parameters to maximize their performance. Actual performance is measured by distributed sensors (both internal and external to the structure) to provide information about functions at all levels (i.e., the sensor model). The difference between the modeled and sensor references provides a residual as the basis for criteria-based digital control. This makes the following functions possible: Enhanced Performance: A functional map of the system makes it possible to predict its performance (in terms of a series of criteria) to meet a very broad range of objectives (rapid response, disturbance rejection, high load, etc.). Condition-Based Maintenance: The difference between the model reference and the sensor reference can be used to identify failure of parts within the system and to plan for repairs and tech mods. Fault Tolerance: Excess resources in the system (in every active component such as motors, brakes, gear trains, etc.) allow the system to maintain operation by isolating the fault and reconfiguring the system to achieve the desired level of performance. The required decision making criteria are unique to the mechanical domain. The requirement for openness in the architecture of the mechanical structure creates the potential for dramatic improvements in cost/performance ratios (as has occurred over the past three decades for computers). A fundamental need is universal operating systems (based on a scientific development of the operational software) in the same manner as demonstrated by Microsoft for personal computers. Because the mechanical system's domain is highly complex and nonlinear, the software must adapt to this complexity. Hence, the science must move towards this opportunity in order to support the development of the general field of intelligent mechanical systems that is just now emerging. To achieve this level of performance will, indeed, be demanding. Whenever we have invested heavily in any major technology (fighter aircraft, nuclear reactors, submarines), we always invest heavily in the

OCR for page 121
Page 131role of the operators and their training in order to maximize the effectiveness of the whole system. This is certainly true for D&D because of the dominant role of the operator. The primary development for the required software is given in the following: 1. Configuration Management This includes complex questions of resource allocation for improved obstacle avoidance, arm stiffness, end-effector dexterity, etc., while prioritizing alternate configurations (tools, power packs, controllers, actuators, etc.) to the decision maker (the D&D system operator). 2. Obstacle Avoidance Develop a mathematical formulation for potential field forces between an array of known obstacles and the structure of the robot. Criteria can be developed to best guide the system through an obstacle-strewn environment (or to find a target) with kinesthetic feedback to the operator through a haptic interface. 3. Criteria Development Criteria are expressed as mathematical descriptions of the physical attributes of the robot (for example: load capacity, energy consumption, accuracy, etc.). Fundamental to the success of this effort is the development of norms for these criteria and their physical meaning to the operator. 4. Criteria Fusion Combinations of these criteria (fusion) become essential in the formulation of task performance indices to enable the decision maker to obtain maximum performance. The operator must be given the opportunity (either through insight or through training) to prioritize (rank) these criteria, to ask for certain combinations to form indices on demand, and to learn what works best for a given task. 5. Fault Tolerance Criteria-based control allows for the selection of the best configuration of the robot to minimize the impact of any given fault (which will usually occur as a reduced load capacity, accuracy, responsiveness, etc., in an actuator) and still maintain operation. 6. Operational Software The required decision making software for management of all available resources is far beyond any standard control approaches (PID, fuzzy logic, adaptive control, etc.). This means that a revolutionary software architecture is required which is not only modular (object oriented) but also extensible, reusable, and operates in real time.

OCR for page 121
Page 132 7. Man-Machine Interface The operation of very complex systems (dual arms, multiple slaves, remote control) to perform demanding tasks (disturbance rejection, handling of ungainly objects) is best achieved by setting operational priorities (selection of criteria, performance indices, threshold levels for fault identification, etc.) by human intervention. Condition-Based Maintenance (CBM). Perhaps the most important development to gain acceptance of remote systems is to assure a high level of availability of that system, to make it rapidly repairable by onsite personnel, and therefore, to substantially reduce forced outage time and the cost of operation. For example, a system that would detect and report that an actuator had degraded to 80% of its performance. Then, the operator (or D&D manager) would evaluate its impact on the spares available at that time and the potential impact on the dismantlement schedule. To accomplish this objective, it becomes necessary to parametrically model each subsystem (all mass, deformation, friction, electrical resistances, etc.), describe its performance (response, load capacity, linearity, etc.) in terms of a selected set of criteria (the hardest problem), establish the effects of their performance on the over-all performance of the larger system, and to embed this decision process partially at the subsystem level (actuators, sensors, communication nodes, etc.) and to download the evaluation results on a disk or directly to the manager's control station. Selecting and defining the physical meaning of a set of performance criteria, both at the subsystem and the system levels, is the most critical issue and would require intense involvement by the D&D community. If it were done well, the operator would be emancipated from the uncertainty associated with maintenance, false alarms would go way down, sudden failures would greatly disappear, and the cost of the overall operation would go down. The project planner would be able to more accurately predict the availability of all the systems, dramatically reducing the threat of major shutdowns and improving safety. The managers would know at all times what systems were at 100 percent, 80 percent, or even 50 percent of their performance levels. They would know what each subsystem's performance meant to what is important to the operation (safety, project schedule, logistics spares, etc.). They would then decide when to maintain the subsystem, how long it would take, should they wait until a scheduled maintenance period, etc. Identification of D&D Task Parameters. In order to define and prioritize a long-term science program for remote systems for use in D&D, it is essential to have a clear parametric definition of the physical tasks to be

OCR for page 121
Page 133performed. This type of data can only be aggregated by careful analysis and by dedicated personnel. The existing needs documents fall far short of the numerical clarity that is necessary to plan a two-decade-long research activity. Detailed, quantitative information could be obtained by in-depth analysis of a few of DOE's most significant D&D tasks (i.e., PNNL—Building 324; Rocky Flats—glove box size reduction; ORNL— Building 3019; and INEEL—Engineering Test Reactor). It would be well to also include some data on the D&D task requirements from our commercial nuclear reactors. Examples include a standard project description; a parametric description of distinct physical tasks for that project; planning experience involving technology transfer, operator training, and cost effectiveness; deployment issues associated with the proposed technology; and comments on the requirements to drive the needed science. The in-depth analysis of the applications chosen should include distinct information that should be provided such as: timeline and frequency of physical tasks duration of each task parameters describing the task: geometry, forces, speeds, accuracy, required dexterity, etc. handling requirements record documentation requirements radiation levels expected level of direct operator involvement

OCR for page 121
Page 134