The synergistic use of data acquired from difference sensors will enable autonomous manufacturing equipment to make faster and more intelligent decisions about the current status of the workspace. Multisensor data fusion deals with mathematical and statistical issues arising from the combination of different sources of sensory information into a single representational format. A fundamental problem in data fusion is associating the data captured by one sensor with that from another sensor or the same sensor at a different point in time. This paper describes a non- statistical unsupervised hierarchical clustering algorithm used to associate the complementary feature vectors extracted from different data sets. Each level in the hierarchy consists of one or more self-organizing feature maps that contain a small number of cluster units based on the combined feature set derived from the original data. The unsupervised learning algorithm ensures that 'similar' feature vectors will be assigned to cluster units that lie in close spatial proximity in the feature map. If the sum- of-square error for the feature vectors associated with a cluster unit is greater than a predefined tolerance, then those vectors are used to create another feature map at the next level of the hierarchy. This growing procedure enables the feature set to control the number of cluster units generated. The hierarchical structure provides an efficient mechanism to deal with uncertainties in correct classification. Experimental studies are present din order to illustrate the robustness of this technique.

Automated tool condition monitoring is an enabling technology in the push to develop fully unmanned machining centers. If this goal can be achieved across a broad range of machine tools, then researchers have assisted industry in moving one step closer to attaining truly flexible manufacturing work cells. Recent advances in the field of image processing technology have led to experimentation with machine vision as a potential means of directly evaluating tool condition. In this work, a machine vision system is employed that permits direct milling inset wear measurement to be accomplished in-cycle. The system is characterized by measurement flexibility, good spatial resolution and high accuracy. The flank wear monitoring system consists of an illumination source, CCD camera and high-resolution microscope lens. the extent of flank wear on the milling inserts was measured using the vision system and an image- processing algorithm. Two vision-based parameters were developed and their efficacy in directly quantifying inset flank were was compared with measurements on a traditional toolmaker's microscope.

With the increased use of extra-fine-pitch (XFP) surface mounted components in electronic products, machine vision techniques for in-process inspection of solder joints during PCB assembly have become highly demanding. Existing algorithms for this purpose are mostly designed to perform inspection based on the analysis of 3D image profiles of individual solder joints. They are computation intensive and often result in expensive implementations. This work introduces a new approach solder joint inspection. With this approach, the inspection problem is considered as a task of detection textural distortions for the 2D proper-view images of a printed circuit board. A real-time texture inspection algorithm based on Fourier Series Analysis is developed to carry out the inspection. The algorithm has been implemented and tested on a PC-based vision system. The results from the tests show that the proposed technique can perform practical PCB inspection successfully. This suggest that real-time and in-process inspection of XFP solder joints may be carried out by a low cost PC-based vision system. In this paper, the principles of Fourier Series Analysis are discussed and the proposed algorithm described. The usefulness of the algorithm in texture inspection in general, and PCB inspection in particular is investigated. The performances of the proposed technique are demonstrated on practical PCB inspection problems.

A low cost, Digital Signal Processor (DSP) based optical position sensor and control system was developed on this work for the vibration control of flexible structures and assembly workcells. This system is based on Texas Instruments' TL220/230 programmable light-to-frequency converters which are interfaces to a TI TMS320C25 DSP. The sensor is a standard 8-pin dip powered by a single DC supply so that it may be mounted on robots or other industrial machines without causing any loading or perturbations. Effects of additive noise are also minimized as the sensor output is pulse position modulated TTL signals.

On-going emphasis on improving quality and reducing waste in polymer refining and converting has resulted in easy to use, cost effective instruments that can provide 100 percent quality analysis of polymer film. A brief tutorial of vision system technology, focusing on the inspection of polymer webs, will be provided by Adaptive Optics Associates Inc., a world leader in high-speed commercial and industrial vision systems. The technique of building polymer-specific 'intelligence' into a quality analysis vision system will be discussed. Included in this presented will be specific examples from field use of the Advisor web inspection as well as the lessons learned during the genesis of this application specific product.

This paper represents a new method to monitor the soldering state of a ball grid array by newly designed digital tomosynthesis system. Firstly, a new digital tomosynthesis (DTS) system, called object-detector synchronous rotation (ODSR), is suggested and designed to acquire images for the soldering state of a ball grid array. Secondary, the shape distortion of DTS images generated by an image intensifier is modeled. And a new synthesis algorithm, which overcomes the limitations of the existing synthesis algorithms is suggested to improve the sharpness of the synthesized image. Also an artifact analysis of the DTS system is performed. Thirdly, the experiment to obtain the cross-sectional images of ball grid arrays is accomplished by the ODSR system.

A reliable servo system design scheme against sensor failures in the speed control loop is proposed. In the proposed reliable control system structure, in addition to the primary output sensor, redundant dissimilar sensors are used to measure different system variables which are more easily accessible and dynamically related to the desired output. In the event of output sensor failure, the measured signal from the redundant sensor can still maintain the system stability and certain performance, such as, tracking ability.

Design of a fuzzy logic controller (FLC) is generally based on trail-and-error because no standard method exists for developing the rule base of an FLC directly from human knowledge. Consequently, the performance of an initial design attempt will, in general, not be satisfactory in terms of certain design criteria such as steady state error and/or oscillatory behavior of the system. In this paper, a method to tune the rule base of such an initial design attempt is presented. The method is based on a novel formulation of an objective function which can then be used in a conventional optimization algorithm. Results of two experiments are reported.

In general, the commuting voltage for a DC pulse width modulator (PWM) fluctuates along with the power supply voltage changes. Therefore, when we describe the mathematical model of a PWM by a first order inertia unit, the giant of PWM unit will vary, e.g. the parameter of current loop will be perturbed. Additionally, in order to raise the safety coefficient of the resist voltage capacity of a large power transistor, the DC supply voltage is usually designed on a low level. As a consequence it causes a low fullness coefficient of starting current in the transient interval. Particularly when the PWM system is in a high velocity setting, the fullness coefficient of starting current is even lower, thus the designed maximum acceleration of system can not be guaranteed, and the transient process slows down. Therefore, when the setting is changed the parameter perturbance of PWM can not be ignored. Using the system identification technique to establish the mathematical model of the PWM system, we also observed that when the load of the positioning servo-system is an eccentric rotating plant, the change of the loading moment along with the angle displacement causes the parameter variation of the PWM system.

This paper presents the results of a low cost, PC-DSP based motion control system development. The Texas Instruments TMS320C31 floating point Digital Signal Processor is selected for the real-time hardware platform while code development and user interface tasks reside on a standard, non real-time PC platform such as the Pentium/Windows 95 systems. An application programming interface has been implemented to facilitate open architecture code development. Finally, application to the control of an Adapt linear robotic workcell is made with 2 input shaper designs serving as the test algorithms.

A method for studying the problem of modeling, identification and analysis of mechanical system dynamic characteristic in view of the impulse response matrix for the purpose of adaptive control is developed here. Two types of the impulse response matrices are considered: (i) on displacement, which describes the space-coupled relationship between vectors of the force and simulated displacement, which describes the space-coupled relationship between vectors of the force and simulated displacement and (ii) on acceleration, which also describes the space-coupled relationship between the vectors of the force and measured acceleration. The idea of identification consists of: (a) the practical obtaining of the impulse response matrix on acceleration by 'impact-response' technique; (b) the modeling and parameter estimation of the each impulse response function on acceleration through the fundamental representation of the impulse response function on displacement as a sum of the damped sine curves applying linear and non-linear least square methods; (c) simulating the impulse provides the additional possibility to calculate masses, damper and spring constants. The damped natural frequencies are used as a priori information and are found through the standard FFT analysis. The problem of double numerical integration is avoided by taking two derivations of the fundamental dynamic model of a mechanical system as linear combination of the mass-damper-spring subsystems. The identified impulse response matrix on displacement represents the dynamic properties of the mechanical system. From the engineering point of view, this matrix can be also understood as a 'dynamic passport' of the mechanical system and can be used for dynamic certification and analysis of the dynamic quality. In addition, the suggested approach mathematically reproduces amplitude-frequency response matrix in a low-frequency band and on zero frequency. This allows the possibility of determining the matrix of the static stiffness due to dynamic testing over the time of 10- 15 minutes. As a practical example, the dynamic properties in view of the impulse and frequency response matrices of the lathe spindle are obtained, identified and investigated. The developed approach for modeling and parameter identification appears promising for a wide range o industrial applications; for example, rotary systems.

This paper describes a market-based approach to controlling a smart matter-based object transport system, in which an array of distributed air jets applies forces to levitate and control the motion of a planar object. In the smart matter regime, the effects of spatial and temporal variation of operating parameters among a multiplicity of sensor, actuators, and controllers make it desirable for a control strategy to exhibit a minimal dependence on system models, and to be able to arbitrate among conflicting goals. A market-based strategy is introduced that aggregates the control requirements of multiple relatively simple local controllers, each of which seeks to optimize the performance of the system within a limited spatial and temporal range. These local controllers act as the market's consumers, and two sets of distributed air jets act as the producers. Experiments are performed comparing the performance of the market-based strategy to a near-optimal model-derived benchmark, as well as to a hand-tuned PD controller. Results indicate that even though the local controllers in the market are not based on a detailed model of the system dynamics, the market is able to effectively approximate the performance of the model-based benchmark. In certain specialized cases, such as tracking a step trajectory, the performance of the market surpasses the performance of the model-based benchmark by balancing the needs of conflicting control goals. A brief overview of the active surface smart matter prototype being developed at Xerox PARC that is the motivation behind this work is also presented.

Open architecture technology is ushering in new advances in the world of computer numerically controlled machines. Yet, some major benefits of open architecture technology have failed to materialized due to the lack of a standard open architecture specification. We propose an open architecture framework to fill the specification void. The proposed framework supports component-based technology by specifying a control class hierarchy, plug-and-play components and a design framework. This framework can be used to build applications ranging from a single-axis device to a multi- arm robot. An example application applying the framework is documented.

Here introduces a concept of Holonic Control in the field of broad sense of control. Nowadays there are many much more complicated things than before in this world, waited to be controlled intelligently for getting a goods' production rate better at a factory or getting things clearer in a complex system system or getting a help in a sense of analyzing a system, etc. In today's control field, for having these complicated things accomplished, we have tried to understand the target system to be controlled clearly, accurately, and precisely. After having gotten this information, it is ready to control for many purposes. But usually this method gives us further complexed problems, more tim consuming because of the size of a system, and comparatively lower robustness. These are used by lack of a flexibility against a sudden change of a system's behavior, giving too much redundant attention to a system, and lack of intelligence. So in order to overcome these problems, here introduces the concept of HOLONIC CONTROL method which always introduces the necessary figures and functions of a system in the whole system to be controlled.

Discrete part manufacturing flows from a design phase in which product information is defined to a manufacturing phase in which the processes are planned and executed. Process planing typically culminates with the generation of numerical control (NC) programs for specific equipment, such as machining centers or turning centers. These NC programs are written in the dialects of the various equipment vendors, for the specific mechanical configuration of the target machine. As a result, porting programs between machines is difficult. Worse, NC programs contain little if any of the product design information. The lack of this information at runtime limits any adaptive control that could direct the process so that final parts more closely conform to the original design.

As production volumes continue to increase and the global market for consumer electronics is getting fiercer, the need for a reliable and essentially fault-free production process is becoming a necessity to survive. The manufacturing processes of today are highly complex and the increasing amount of process data produced in making it hard to unravel the useful information extracted from a huge data set. We have used multivariate and nonlinear process modeling to examine the surface mount production process in a high volume manufacturing of mobile telephones and made an artificial neural network model of the process. As input parameters to the model we have used process data logged by an automatic test equipment and the result variables come from an Automatic Inspection system placed after the board manufacturing process. Using multivariate process modeling has enabled us to identify parameters, which contributes heavily to the quality of the product and can further be implemented to optimize the manufacturing process for system production faults.

Up-to-date manufacturing equipments of production of rotational parts in small series are lathe-centers and CNC grinding machines with high concentration of manufacturing operations. By the use of these machine tools it can be produced parts with requirements of increased accuracy and surface quality. In the lathe centers, which contain the manufacturing procedures of lathes using stationary tools and of drilling-milling machine tools using rotational tools, non-rotational surfaces of rotational parts can also be produced. The high concentration of manufacturing operations makes necessary the planning and programing of the measuring, monitoring and quality control into the technological process during manufacturing operation. In this way, taking into consideration the technological possibilities of lathe canters, the scope of computer aided technological planning duties significantly increases. It is trivial requirement to give only once the descriptions of the prefabricated parts and ready made parts. Starting taking into account these careful considerations we have been developing the planning system of technology of body of revolution on the base of GTIPROG/EC system which useful for programming of lathe centers. Out paper deals with the results of development and the occurring problems.

Nowadays developing of the new technologies which decrease the unfavorable environmental effects of the traditional manufacturing process and using them in the industrial production get even more into the focus. The development of environmentally clean manufacturing technologies require careful and circumstantial analysis. In this paper the substantial measurements performed during experiments of environment-friendly drilling processes are shown including the examination of new and worn tool geometry, the geometry of the machined surfaces, the change of feed force and torque as well as the temperature of the contact zone between work and too, an last but not least the analysis of the modified vibration. The mathematical model based on the known solvation method of partial differential equation is a good base model which can further be developed. Nowadays a more general distributed parameter model that included both transverse and torsional vibration of the drill bit also seems desirable. The effects of damping and temperature must also be investigated.

The intelligent automatization of the production of worm driving pairs has become a hot issue because of the requirements of the up-to-date high-grade production. In present production practice home and abroad only some elements have been realized, but a general system has not been developed. Our paper presents the structure of a general system and some considerations and methods for the realization of some of its elements and subsystems. Some elements of the system we have already produces but the rest of the elements and their connection into the system are being worked out nowadays as our main research subject. We think that both the theoretical basis and methods, and the up-to-date equipment developed for the realization, increase the productivity and the quality of the final product.

This paper describes the analysis, design and control of a novel, single-phase motor with a unique behavior resulting from the use of rotating power electronics mounted to the motor armature. Coils on the armature are selectively shorted by power MOSFET's which rotate with the armature, and torque is produced by interaction between currents induced in the shorted coils and the magnetic field produced by a stationary field coil. Control is limited to the timing of which armature coils are to be shorted as a function of armature speed and angle, it is possible to modulate torque production and obtain torque or speed control using only single-phase ac power and without the use of brushes or permanent magnets. An electro-mechanical model for this type of motor is presented and validated with respect to experimentation. The results show promise for achieving low- cost, adjustable-speed drives using this novel method of rotating electronics, optical communications, and computed commutation.

In this work, the problem of vibration control for a contactless magnetic leadscrew system is considered. A contactless drive system is a magnetic nut/leadscrew and air bearing assembly that operates on the principle of magnetic/aerodynamic suspension to position a load with high accuracy. However, the dynamics of such system is lightly damped, load dependent, and generally difficult to stabilize by conventional linear controllers. Therefore, the technique of recurrent neural network is applied to separate the oscillatory signals so that passband shaping can be carried out to regulate plant dynamics and to reject disturbances. This controller possesses a modular structure and is easy to implement. Experimental results also confirm the vibration suppression capabilities of this controller.

Applying the principle of mechatronics, this paper presents design considerations for a new type of sensor-embedded 'smart' long cane which can serve as an orientation and travel aid for the blind. The smart long cane uses a set of miniaturized ultrasonic sensors to detect obstacles along the travel path, and provide human voice feedback on the obstacle's height and distance to guide the cane user away for head collisions and subsequent injuries. Topics discussed include sensor array design, placement, and integration, structure of the embedded software, and ergonomic issues for the new cane design.

The advent of more complex mechatronic systems in industry has introduced new opportunities for entry-level and practicing engineers. Today, a select group of engineers are reaching out to be more knowledgeable in a wide variety of technical areas, both mechanical and electrical. A new curriculum in mechatronics developed at Virginia Tech is starting to bring students from both the mechanical and electrical engineering departments together, providing them wit an integrated perspective on electromechanical technologies and design. The course is cross-listed and team-taught by faculty from both departments. Students from different majors are grouped together throughout the course, each group containing at least one mechanical and one electrical engineering student. This gives group members the ability to learn from one another while working on labs and projects.

To assess the working condition of a rolling element bearing, the condition monitoring system should be located as close as possible to the bearing to take advantage of shorter signal transmission path, increased signal-to-noise ratio, and reduced complexity of the signal processing electronics. The advantages of integrated sensing are presented in this paper, with a focus on the design and analysis of a miniaturized sensor module. Mechatronic principles have been applied to treat the various subjects in a synergistic way. To complement analytical studies, experiments have been conducted on a scaled-up version of the sensor module to analyze the system dynamic response. The result obtained provided insight into the electromechanical interaction within the module as well as input for the system implementation using miniaturization technologies.

Increasing demands for product safety and reliability requires the development and implementation of innovative condition monitoring mechanisms that are an integral part of the system to be monitored. Since an advanced condition monitoring system often consists of a variety of sensing, controlling, and actuating components, their effective and efficient integration requires the application of mechatronic design principles to achieve the desired synergy. This paper presents several aspects related to the design and implementation of a sensor-embedded mechatronic bearing, which can be used for the condition monitoring of various critical machine systems.

The design of real-time control software for a mechatronic system must be effectively integrated with the system hardware in order to achieve useful qualitative benefits beyond basic functionality. The sought-after benefits include: rapid development, flexibility, maintainability, extensively, and reusability. In this work we focus upon the interface between the device drivers and the control software with the aim to properly design this interface to best realize the aforementioned benefits. The results of this fundamental research include the development of an easily manageable set of four C++ object classes following an object-oriented approach to software design. These Universal Mechatronic Objects (UMOs) are applicable to a wide spectrum of actuators including dc motors, stepper motors, and solenoids; and sensors including pressure sensors, microswitches, and encoders. UMOs encapsulate the interface between the electrical subsystem and the control subsystem, providing the control software developer with a powerful abstraction that facilitates the development of hardware-independent control code and providing the electrical subsystem developer with an effective abstraction that facilitates the development of application-independent device drivers. Objects which are intuitively related to hardware components of the mechatronic system can be declared using the UMOs early in the system development process to facilitate the rapid concurrent development of both the electrical and the control subsystems. Our UMOs were developed as part of a project to implement a real-time control system for a z-theta robotic manipulator. The z- theta manipulator is one component of the Minifactory project in the Microdynamic Systems Laboratory at Carnegie Mellon University. The goals of this agile assembly project include the reduction of factory setup and changeover times, plug-and-play type modularity, and the reuse of its components. The application of UMOs to the manipulator software development is shown to be consistent with these goals.

To develop a next-generation manufacturing system with flexibility, it is necessary to define H/W and S/W architecture based on an open architecture concept. To satisfy the open architecture concept, in this paper, H/W is divided into three parts for hard real-time task, soft real- time task, and non-real-task. And, S/W also is divided into three layered level for application level, system level, and device level. Due to the modulization of H/W and S/W, it is easily possible to modify and/or upgrade the control system including S/W and H/W according to user requirements. Especially, it is also possible to develop a new control system by modifying a part of application and device level. To show the validity of the proposed method, a detail control concept of wire bonder with the open architecture concept is described.

This work addresses the design of optimal input shaper and its application to the motion control of a robotic workcell. An optimal shaper is proposed to tradeoff performance and robustness according to assembly specifications of the workcell. The optimal shaper, along with standard shaper designs such as Zero Vibration, zero vibration and derivative, and extra insensitive are applied to conduct cycle time testing on a two-axis adept technology robotic workcell. The performance for various unknown loading conditions is observed. It is shown that the optimal shaper produces the best overall results.

Computer vision systems have been used in recent years to perform automated antibiotic susceptibility test based on the disk-diffusion method. However, certain organisms do not reflect light very well. As such, the reliability of such automated inspection systems is sometimes not as high as expected. This paper proposes to use texture analysis to improve the quality of test images and thereby simplify the inspection tasks. Adaptive texture filters are used to maximize the difference between regions of interest in a test image and the background, enabling a thresholding operation to be carried out easily. The principles of adaptive filtering for texture analysis are discussed. A training algorithm is presented to generate optimized filters for generic texture inspection problems. An experimental study is carried out to investigate the performance of this technique in highlighting poorly reflecting organisms in antibiotic susceptibility testes.

This paper presents a new class of solution to the inverse kinematics for a modular arm with N-active links. Due to the design advantage of the active link, we bring the inverse kinematics problem from a 3D space to a 2D space. Based on the proposed methodology, a class of simple solutions can be quickly and easily obtained. Simulation results prove the validity of the solutions. Results with real modular arm are also included in this paper.