Detailed simulation of manufacturing processes is traditionally a time intensive and computationally expensive process. This research is aimed at reducing the time for a single cycle simulation by using a fuzzy model to describe the process. The training set is generated by a designed experiment and the fuzzy parameters are optimized by backpropagation.

This paper introduces a new tool breakage detection technique in face milling by using an unsupervised neural network. The cutting force signals are modeled by an autoregressive (AR) model where parameters are estimated recursively at each sampling instant using a parameter adaptation algorithm based on a RLS (Recursive Least Square). Experiment indicates that AR parameters are good features for tool breakage, therefore it can be detected by tracking the evolution of the AR parameters during machining. ART 2 (Adaptive Resonance Theory 2) neural network is used for clustering of tool state using these parameters, and this network has a self organized capability without supervised learning. Therefore, this system operates successfully without a priori knowledge of the cutting process.

This paper describes an algorithm for determining the optimal placement of a robotic manipulator within a workcell for minimum time coordinated motion. The algorithm uses a simple principle of coordinated motion to estimate the time of a joint interpolated motion. Specifically, the coordinated motion profile is limited by the slowest axis. Two and six degrees of freedom examples are presented. In experimental tests on a FANUC S-800 arm, the optimal placement of the robot can improve the cycle time of a robotic operation by as much as 25%. In high volume processes where the robot motion is currently the limiting factor, this increased throughput can result in substantial cost savings.

Taking account of different influences of pressure, rocker amplitude and rocker center angle upon polishing quality for optical lens, a fuzzy self-learning algorithm based on semantic inference is developed and a prediction system is built up with the adjustment of membership function. The introduction of the intellective control makes the polishing process be less influenced by other technological parameters than the conventional method.

A novel model for CMM measurement process using the concept of Self Avoiding Walk is presented. The objective is to find out a sampling lattice which results in reduced measurement time without affecting the precision of the measurement. Sampling in honeycomb lattice is shown to have less mean square displacement than sampling in square lattice as a result of which sampling in honeycomb lattice requires less measurement time than sampling in square lattice. Further it is shown that the sampling in honeycomb lattice resembles the unaligned systematic sampling which again increases the precision of the sampling in honeycomb lattice.

Flow shop production lines are very common in manufacturing systems such as car assemblies, manufacturing of electronic circuits, etc. In this paper, a systematic procedure is given for generating event-timing equations directly from the machine interconnections for a generalized flow shop system. The events considered here correspond to completion times of machine operations. It is assumed that the scheduling policy is cyclic (periodic). For a given flow shop system, the open connection dynamics of the machines are derived first. Then interconnection matrices characterizing the routing of parts in the system are obtained from the given system configuration. The open connection dynamics of the machines and the interconnection matrices are then combined together to obtain the overall system dynamics given by an equation of the form X(k+1) equals A(k)X(k) B(k)V(k+1) defined over the max-plus algebra. Here the state X(k) is the vector of completion times and V(k+1) is an external input vector consisting of the arrival times of parts. It is shown that if the machines are numbered in an appropriate way and the states are selected according to certain rules, the matrix A(k) will be in a special (canonical) form. The model obtained here is useful or the analysis of system behavior and for carrying out simulations. In particular, the canonical form of A(k) enables one to study system bottlenecks and the minimal cycle time during steady-state operation. The approach presented in this paper is believed to be more straightforward compared to existing max-plus algebra formulations of flow shop systems. In particular, three advantages of the proposed approach are: (1) it yields timing equations directly from the system configuration and hence there is no need to first derive a Petri net or a digraph equivalent of the system; (2) a change in the system configuration only affects the interconnection matrices and hence does not require rederiving the entire set of equations; (3) the system model is easily put into code using existing software packages such as MATLAB.

A new matrix formulation is given that allows fast, direct design and reconfiguration of rule- based controllers for manufacturing systems. Given a bill of materials or assembly tree, Steward's sequencing matrix is constructed. Then, resources and agents are added through `resource matrices' such as those used by Kusiak, and extra inputs are added to resolve shared-resource conflicts. The result is a multiloop DE controller with outer loops for dispatching of shared resources. The matrix formulation allows a rigorous analysis of deadlock in terms of circular blockings, siphons, and the numbers of resources available; this allows efficient dispatching and routing with deadlock avoidance. An assembly task is used to illustrate the concepts introduced.

To acquire the complete surface description of a nontrivial object using range cameras several range images from different viewpoints are needed. We present a complete system to automatically acquire a surface model of an arbitrary part and outline the constraints this system places on a solution to the problem of where to position the range camera to take the next range image, i.e. the next best view (NBV) problem. We present a solution which uses no a-priori knowledge about the part and which addresses the most crucial of these constraints which is that each new range image must contain range data of part of the object's surface already scanned so that it can be registered with the previously taken range images. A novel representation, positional space, is presented which is capable of representing both those hypothetical sampling directions which could scan the unseen portions of the viewing volume and those which could rescan parts of the object. In addition, positional space makes explicit the actual sampling directions available given a particular range camera and the set of relative motions possible between it and the object. A solution of the NBV problem is achieved by aligning the positional space representation of the range camera with the positional space representations of the scanned portions of the object and the unseen portions of the viewing volume using simple translations. Since complex motions of the range camera in its workspace are represented by translations in positional space the search for the next best view is computationally inexpensive. No assumptions are made about the geometry or topology of the object being scanned. Occlusions and impossible sensing configurations are easily addressed within this framework. The algorithm is complete in the sense that all surfaces that can be scanned will be scanned. In addition, confidence values for range samples can be used to instruct the algorithm to position the range camera so that all surfaces of the object are scanned with at least a minimum confidence wherever possible. The algorithm can determined when all scannable surfaces have been sampled and can be used with any range camera provided a model of its exists. The algorithm can also accommodate nearly any set of relative motions possible between the range camera and the object.

Much of the early work in robotics focused on developing guaranteed plans for accomplishing tasks specified at a high level. Such task specifications might be of the form `mesh these two gears', or `place part A inside region B'. It is now always possible, however, especially in the realm of assembly planning, to generate guaranteed plans. For example, errors in tolerancing of the parts might render an assembly infeasible. The Error Detection and Recovery (EDR) framework of Donald was developed to deal with these inadequacies of the guaranteed planning framework. EDR strategies will either achieve a goal if it is recognizably reachable, or signal failure. Given a geometrically-specified goal region G, an EDR strategy involves computing a failure region H and a motion plane that will terminate recognizably either in G or H. The question addressed in this work is that of computing sensing strategies for distinguishing which of G and H have been attained. We propose a method with which we can strengthen the guarantee of reaching G or H into a guarantee of recognizability. In particular, we show how to configure a sensor or set of sensors so that a target object in G or in H can be distinguished. Our approach assumes a general sensor model, and builds on algorithms for computing partial visibility maps based on point-to-point visibility between objects in an environment. We characterize recognizability and confusability regions, that is, sensor placement regions from which an object in G or in H can be distinguished, and regions from which attainment of G or H could be confused.

In this paper, we consider a flexible manufacturing system which consists of one reliable machine and produces two distinct products simultaneously. The machine is assumed to be flexible, i.e. there is no setup time or cost incurred when the machine switches from one product to the other. Therefore, it is possible to switch production as often as necessary between products to achieve any desired production mix within the capacity set. The processing times for individual products are constant and order of magnitude smaller than the planning horizon. We formulate an optimal control problem to regulate the production flow tracing a given set of demand. The problem is then solved exactly using Pontryagin's Minimum Principle.

Strategies for managing data flow are described for a system which controls processes to compensate machine tool and process errors. A CAD representation of the part is the basis for generating the data used in the processes and analyses. A process-oriented view of part features is used for tracking data regarding design, part programming, manufacture, error compensation, and inspection. Part ID and file name nomenclature have been developed to facilitate data identification. All data are stored in a database, enabling error trends over time to be studied so that error compensation algorithms may be refined.

This paper studies the cutting tool wear with fractal geometry. The wearing image of the flank has been collected by machine vision which consists of CCD camera and personal computer. After being processed by means of preserving smoothing, binary making and edge extracting, the clear boundary enclosing the worn area has been obtained. The fractal dimension of the worn surface is calculated by the methods called `Slit Island' and `Profile'. The experiments and calciating give the conclusion that the worn surface is enclosed by a irregular boundary curve with some fractal dimension and characteristics of self-similarity. Furthermore, the relation between the cutting velocity and the fractal dimension of the worn region has been submitted. This paper presents a series of methods for processing and analyzing the fractal information in the blank wear, which can be applied to research the projective relation between the fractal structure and the wear state, and establish the fractal model of the cutting tool wear.

In this paper the deformation of the two typical built up ball leadscrew joints under a load is discussed in a deep-going way. A theoretical basis for the right design, manufacture, assembly and use of the built up precision ball leadscrew is provided.

The screw-thread steel bar is a special rolled-one, it moves at a speed of 7 - 12 m/s on production line. In laboratory, it is difficult to reappear it truly. However, according to proportional simulation method, increase the camera's exposure time when the steel bar runs at a lower speed, then we can get image just as the one taken at the workroom; to capture necessary information, structural illumination is adopted, and sample from different angle; with the help of CCD camera's special time delay integration technique, it is easy to process this kind of motion-degraded images; using Soble operator to extract edge will greatly reduce data; the adaptive template-matching is employed to realize feature recognition; by making up straight lines with least mean square method, can work out the useful parameters such as the internal and external diameter, the distance of cross rib and the height of longitudinal rib.

In recent years, manufacturers of high precision mechanical parts have been required to produce increasingly complex designs, in smaller lot sizes, with improved quality. These requirements demand lower process costs, shorter development cycles and more accurate manufacturing technologies. To meet these demands, manufacturers are attempting to both improve process quality and provide better CAD/CAM integration. The technique of on- machine acceptance provides one mechanisms for improving the part inspection and verification process. This approach allows one machine and one process capability model to be used for both fabrication and inspection, reducing capital cost and overall cycle time. However, the on-machine acceptance technique possesses greater potential than as simply an alternative mechanism for verifying part geometry. If the inspection capability information generated by on-machine acceptance processes can be made available to designers, it can be used to create a design-for-inspectability environment and help realize the benefits of concurrent engineering. This paper proposes a novel architecture which integrates on-machine acceptance with an agent-based concurrent design environment, for reducing both the cost and production time for high quality, small lot size, mechanical parts. This work has focused on the production of stainless steel pressure vessels at the Integrated Manufacturing Technology Laboratory manufacturing cell, located at Sandia National Laboratories, California.

A boiler system is an integral component of a thermal power plant, and control of the water level in the drum of the boiler system is a critical operational consideration. For the drum level control, a 3-element proportional-integral-derivative (PID) control is a popular conventional approach. This scheme works satisfactorily in the absence of any process disturbances. However, when there are significant process disturbances, the 3-element PID control scheme does not perform well because of lack of knowledge of proper controller gains to cope with such disturbances. Inevitably over time and use, PID controllers get detuned. Hence, there is good motivation to investigate alternatives to this control scheme. Multivariable control of drum boiler systems has been studied by many researchers. However, these approaches assume some process model equations (to a more or less extent) to design a controller. This paper presents a model-free approach in the sense that no plant equations are assumed. Only data is used to gain knowledge about the process, and the performance of the existing PID control scheme is observed. Based on this process knowledge, an intelligent control technique is developed, (artificial) neural network control (NNC). The technique proposed in this paper was tested on a process simulator. This paper shows that an intelligent control scheme such as NNC gives better performance in rejecting process disturbances when compared to 3-element PID control scheme.

The paper addresses the problem of grasping moving objects from a vibratory feeder with robotic hand-eye coordination. Since the dynamics of moving targets on the vibratory feeder are highly nonlinear and impractical to model accurately, the problem has been formulated in the context of Prey Capture with the robot as a `pursuer' and a moving object as a passive `prey.' A hierarchical vision-based intelligent controller has been developed and implemented in the Factory-of-the Future Kitting Cell at Georgia Tech. The first and second levels are based on the principle of fuzzy logic to help the robot search for an object of interest and then pursue it. The third level is based on a backpropagation neural network to predict its position at which the robot gripper grasps it. The feasibility of the concept and the control strategies was verified by two experiments. The first experiment showed that the fuzzy logic controller could command the robot to successfully follow the highly nonlinear motion of a moving object and approach its vicinity. The second experiment demonstrated that the neural network could estimate its position fairly accurately in a finite period of time after the command of grasp operations was issued.

Currently, un-round process (such as piston in internal-combustion engine etc.) is mainly realized by hard contact turning and ellipse milling. This paper introduces the computer-aided control in non-round process, using a high speed voice coil motor which can move to and fro in straight line to drive the cutting tool directly and so as to realize the un-round process under the control of computer. The precision can reach 1 micrometer. It exemplifies the process on convex varying elliptical piston. The best process curve is deduced. Double-CPUs are adopted to realize high speed detection and curve imitation in on-line process. IGBT is selected to drive the motor, optics grid decoder with the resolution of 0.25 micrometer for position detection. The self-adaptive control and feed forward control of sliding model on varying structure assures the request on system's dynamic response and stability. The simulation results reach the expected goal of system design.

The Internet offers tremendous potential for rapid development of mechanical products to meet global competition. In the past several years, a number of geometric algorithms have been developed to evaluate manufacturing properties such as feedability, fixturability, assemblability, etc. This class of algorithms is sometimes termed `DFX: Design for X'. One problem is that most of these algorithms are tailored to a particular CAD system and format and so have not been widely tested by industry. the World Wide Web may offer a solution: its simple interface language may become a de facto standard for the exchange of geometric data. In this preliminary paper we describe one model for remote analysis of CAD models that we believe holds promise for use in industry (e.g. during the design cycle) and in research (e.g. to encourage verification of results).

The National Institute of Standards and Technology has developed a modular definition of components for machine control, and a specification to their interfaces, with broad application to robots, machine tools, and coordinate measuring machines. These components include individual axis control, coordinate trajectory generation, discrete input/output, language interpretation, and task planning and execution. The intent of the specification is to support interoperability of components provided by independent vendors. NIST has installed a machine tool controller based on these interfaces on a 4-axis horizontal machining center at the Pontiac Powertrain Division of General Motors. The intent of this system is to validate that the interfaces are comprehensive enough to serve a demanding application, and to demonstrate several key concepts of open architecture controllers: component interoperability, controller scalability, and function extension. In particular, the GM-NIST Enhanced Machine Controller demonstrates interoperability of motion control hardware, scalability across computing platforms, and extensibility via user-defined graphical user interfaces. An important benefit of platform scalability is the ease with which the developers could test the controller in simulation before site installation. The EMC specifications are serving a larger goal of driving the development of true industry standards that will ultimately benefit users of machine tools, robots, and coordinate measuring machines. To this end, a consortium has been established and cooperative participation with the Department of Energy TEAM program and the US Air Force Title III program has been undertaken.

Manufacturing is a challenging activity. One must coordinate many activities to achieve success. There appears to be no magic formula which ensure quality. Simple prescriptions for all of manufacturing ills have been suggested, but the theory works better than the practice. This paper explores manufacturing from the standpoint of the interactions of workers, management, and the technology they use in their jobs. These three factors form a complex system, and to optimize the system is virtually impossible without a greater level of understanding. Technology is clearly one factor which makes a company excel, but it is not the only factor. Technology cannot be looked upon as the savior of manufacturing, but as one component of a complex system.

In manufacturing supply networks today, companies are separated by boundaries through which little or no production information flows. Existing production scheduling and inventory control paradigms are not efficient for supply networks of autonomous competing companies, because production information is proprietary and control is sovereign. What is needed to achieve production optimization across the network is a new paradigm for operational control, based on the exchange of Just-Enough-Information (JEI). We have invented a Demand- Availability-Order (DAO) algorithm that creates a `pull' system distributed recursively across computers in a supply network, with each company using only local information. We believe that DAO is the key to JEI. Practical application depends on generalizing DAO and on complementing it with a broad range of related systems.

As many main components of optical devices and measurement equipment, as lenses, mirror, and targets, are made of glasses and ceramic materials, the integrity of these components is an important issue for quality prediction and assessment. It is well known, that machining processes can greatly influence the quality of products, not only at surfaces but also directly beneath it. This problem is even more severe, when machining hard and brittle materials as ceramics and glasses, as the chip removal process can lead to intensive crack initiation and stresses at surfaces, and thereby to sudden failure when loading those components. But under certain machining conditions, a small range of plastic flow has been observed in some ceramics and glasses as well. This change in their behavior is known as the brittle-to-ductile transition phenomenon. Until now, the transition process itself is not understood sufficiently, there is still no conclusive description of the process. The prediction of the critical depth of cut or the maximum pressure before crack formation is one of the key issues for maximizing the material removal rate and assuring the reliability of components at the same time. Central to the scientific evaluation of the micro cracking in machining of brittle solids is the indentation test, today widely adopted as material `hardness' indicator. In engineering, the indentation test is also considered as a simplified model of the penetration of abrasives into work material during the machining process. Simulations for silicon have confirmed the dependency of crack initiation on the indenter tip radius, demonstrating its brittleness in addition to elastic response of the bulk and to deformation around the indenter. These first results represent an important step towards the simulation of the brittle-to-ductile transition, which will greatly enhance the understanding of machining of hard and brittle materials. The results will then in turn form the basis for further improvement in machining technology and higher reliability of ceramic components by controlled surface integrity.

We discuss how the combination of a realistic human figure with a high-level behavioral control interface allow the construction of detailed simulations of humans performing manual tasks from which inferences about human performance requirements can be made. The Jack human modeling environment facilitates the real-time simulation of humans performing sequences of tasks such as walking, lifting, reaching, and grasping in a complex simulation environment. Analysis capabilities include strength, reachability, and visibility; moreover results from these tests can affect an unfolding simulation.

In this paper, a method for performing time-scaled distributed simulation is described for control of heterarchical manufacturing systems, where it is used as a mechanism by which highly autonomous entities in these systems can cooperate implicitly and to adapt to disturbance events in real-time. A replica of the system software, executing at a speed that is several orders of magnitude faster than real time, serves as an accurate, detailed, distributed simulation model that can be developed at low cost. Performance estimates obtained from simulations are used as feedback by entities in the system to iteratively improve global coherence. The time-scaled approach eliminates the need for explicitly synchronizing events and thereby eliminates the complexity associated with discrete event distributed simulation approaches. The method is applied in the paper to a manufacturing control system in which control laws with a continuous feedback structure are embedded in the autonomous workpiece entities and used to adjust the timing of events required to produce workpieces. The control laws ensure the stability and convergence of the system, while allowing it to be effectively controlled with minimal global information.

We describe an implementation of the simulation of human grasping for manufacturing tasks using a realistic human figure with a high-level behavioral control interface. Grasping simulations can then inform the manufacturer as to requirements on human grasping in manufacturing environments. We further describe the integration of a reach planner with grasping to provide simulation of reaching and grasping tasks. The simulation of human grasping proceeds from work in cognitive science, psychology, and robotics, in which human grasping is described as a primarily tactile activity. Information about target geometry is derived from contact with the object to be grasped, and the simulation is driven by this collision information. Reaching and grasping have been integrated to provide more automated control of the simulation of grasping. A motion planning approach controls human arm movement: given a position in 3D space, the reaching algorithm automatically computes a collision-free and strength-feasible motion sequence to move the hand to the desired position. If a tool is held in the hand, the collision of the tool with the environment can also be avoided.