Chaos theory is used to conceptualize a methodology to control cutting tool chatter in the turning process. The underlying information contained in a Mandelbrot set -- developed from an existing nonlinear model of cutting tool dynamics -- is supposed to be captured by a multilayer feed forward neural network. The neural network may be used as a control parameter variation selector in a control loop. We anticipate real-time implementation of the proposed methodology.

An optical telemetering system which measures the temperature of the rotating spindle of a machine is proposed. Our system uses an optical data transmission to realize high-speed precise telemetering, it is robust for the external electric-magnetic field. Moreover it can measure the temperatures at eight different positions simultaneously and the data is transmitted with only two pairs of infrared LED and photodiode by using an asynchronous serial data transmission. We have made an experimental system and measured some fundamental characteristics. The time required for the data transmission is 48 microseconds a channel. It is confirmed experimentally that the accuracy in temperature measurement is 0.4 degrees Celsius.

In this paper, vibration monitoring and measurement carried out in the newly developed nanometer metrology laboratory in Birmingham University, is described with respect to measurement methods and instrument performance. Two types of instrument -- piezoelectric accelerometer (B&K type 8318 with a type 2140 bus-controlled frequency analyzer) and laser interferometer (Renishaw ML10 laser interferometer with the Keithley multi-channel FFT analyzer) were used in vibration measurement on capability verification of a vibration isolation system. Vibration results from a concrete block are presented. From the point of view of the measured vibration results, it is demonstrated that the performance of the above two instruments is not completely the same in the different frequency ranges. The related comparison and discussions are presented in this paper.

Modern industry increasingly demands that machine tools operate continuously as elements within computer integrated manufacturing cells. Accurate monitoring of machine condition is the key to predicting failures that would result in quality defects or costly unplanned production stoppages. This paper presents research into the use of a novel rotary motion sensor for the condition monitoring of rolling element bearings. This sensor, the incremental motion encoder (IME), is based upon a patented development of the optical encoder commonly used in machine tools for sensing angular position and rotational speed. The IME combines these measurements with that of shaft center position in two dimensions. This motion of the shaft center is directly related to the condition of the bearings supporting the shaft. To illustrate the IME principle experimental results showing the ability of the sensor to distinguish between bearing defects and external sources of vibration are presented. Measurement of shaft loading with the sensor is also described. Currently machine tool condition is most often measured by external sensors, such as accelerometers or acoustic emission transducers, which are not parts of the machine itself. The IME is ideally suited to being designed into a machine tool so as to integrate a condition monitoring facility into the computer control of the machine. The paper concludes by describing the current technology which allows sensors based on the IME principle to be integrated directly into rolling element bearings for this purpose.

During metal-removal processes, acoustic emission energy variation in certain frequency bands is correlated with tool wear. Energy estimation in real-time, therefore, is potentially important in the monitoring and control of machine processes. In practice, a commonly used energy detector is simply a linear, time-invariant filter followed by a magnitude-square operator. Such a conventional energy estimation technique is limited by an unavoidable time- frequency resolution trade-off due to properties of the linear time-invariant filter. Furthermore, the result of this estimation is sometimes physically uninterpretable. In this paper, we describe desired properties for a quadratic acoustic emission energy detector, then formulate an alternative energy measure which is more generally valid than the conventional technique. It is also more flexible than quadratic systems based on Teager's energy operator. The features extracted by the quadratic detector, such as position and energy of the main resonance, preserve the details, in both time and frequency, which are useful for machine tool monitoring. Examples of relative performance are demonstrated on both synthetic signals and on data from metal drilling applications. We conclude that the generalized quadratic energy detector provides superior energy estimation.

This paper addresses our ongoing funded research 'R&D on Intelligent Integrated Numerical Control (I²NC) System.' In the paper, the intelligent characteristic of intelligent numerical control (INC) system is analyzed and the architecture of an INC system is proposed. The system's hardware configuration and software design are presented. Finally, the hierarchical system's three intelligent levels are also discussed.

This paper presents the use of nonlinear optimal predictive control on rubber artificial muscles. A neural network which can learn dynamic characteristics of rubber artificial muscles is employed in this control scheme in order to obtain the model. The actuator has strong nonlinear characteristics due to compression of both rubber and air. It is difficult to control with precision using the linear control theory. A suitable control method for this actuator has yet to be established. Computer simulations of a common pneumatic actuator, which is modeled quite easily, were carried out to verify the proposed control scheme. This control scheme was adapted to a practical one dimensional robot manipulator which was developed for this study. The theory was compared with PID control in order to explain availability of the theory. When desired targets are changed as a step response, this actuator can be controlled precisely in these simulations and experiments. The actuator can be controlled adaptively under changes of characteristics if the learning of the neural network continues. It is obvious from these results that the proposed control scheme is effective for rubber artificial muscle.

An automated and portable weigh-in-motion system has been developed at Oak Ridge National Laboratory for the purpose of manifesting cargo onto aircraft. The system has an accuracy range of plus or minus 3.0% to plus or minus 6.0% measuring gross vehicle weight and locating the center of balance of moving vehicles at speeds of 1 to 5 mph. This paper reviews the control/user interface system and weight determination algorithm developed to acquire, process, and interpret multiple sensor inputs. The development effort resulted in a self- zeroing, user-friendly system capable of weighing a wide range of vehicles in any random order. The control system is based on the STANDARD (STD) bus and incorporates custom- designed data acquisition and sensor fusion hardware controlled by a personal computer (PC) based single-board computer. The user interface is written in the 'C' language to display number of axles, axle weight, axle spacing, gross weight, and center of balance. The weighing algorithm developed functions with any linear weight sensor and a set of four axle switches per sensor.

We have developed a new type of scanning in-line shape-matching sensor with a miniature optical scanner based on a simple shape-matching algorithm. The sensor consists of a sensor head, a controller, an interface board and CPU. The sensor head, with dimensions of 30 by 40 by 40 mm³, consists of a miniature optical scanner, a laser diode with a collimator lens, a photo detector and an amplifier circuit. The controller consists of signal processing circuits which generate the scanning position information from the drive circuit of the scanner. The interface board inputs binary level signals to the CPU. The CPU is used for the calculation and judgement of the obtained data and for displaying the results of its output. The shape- matching is performed by comparing the obtained scanning data to the pre-inputted original data. The original data is inputted to the memory by scanning the beam on the correct status object. If both sets of data completely agree, it defines that detected object as equal to the original object. For comparison, the most typical parameter of the data, such as the pulse width or the number of pulses, is adapted. For example, in the case of size difference between screws, each pulse width of scanned data is changed. As a result, size detection can be achieved. As this scanning type sensor, not a CCD camera type, has features such as no external lighting, lower cost and miniature size, it could be applied in the area of simplified shape matching on the factor assembly line.

Due to the inexact science of grinding, a technique was desired to automate the grinding process in order to complete a finished profile with a closed loop control. A desired finished shape, or profile, is input to the system. The closed loop control plans the optimal grinding process, and drives the machine controllers to complete the task. This technique can be applied to many types of surface grinding applications such as optical grinding, precision surface grinding for coordinate measuring systems, complex curve grinding, as well as many others. The example illustrated is the closed loop application of railway rail grinding. Vista (vision based transverse analyzer) was intended as a solution to the difficult problem of railroad rail grinding. Due to complexities in analysis, and measurement limitations, rail grinding had evolved as an inexact art. Vista was created to refine the analysis techniques, and to automate the process of rail grinding. The intent was to design and produce a system fully capable of grinding railroad rails efficiently without the aid of human intervention. Vista is comprised of two parts; a rail measurement system (RMS), and an expert system. These devices become the host controller of present day rail grinding machinery, increasing efficiency and accuracy in the rail grinding environment.

The paper presents a wheel chair system with the capability of self-localization and obstacle avoidance. Firstly, the approaches of landmark recognition and the self-localization of the wheel chair are described. Then, the principal of the obstacle avoidance using a laser range finder is described. Subsequently, the total system of the wheel chair is introduced. Finally, a navigation experiment is given. Experimental results indicate the effectiveness of our system.

The article presents laboratory equipment that contains information about the installation of oil pipes and righting-lifters. Material based on the system provides useful information about installation, such as the quantity of oil pipes and righting-lifters being used, and whether all of them satisfy the standard. The information is valuable both for examination and maintenance.

As one of the interesting optical techniques for measurements of velocity, the spatial filtering method is treated briefly in this paper. We show theoretical analysis and calculation of spatial filtering velocimetry, and discuss two-dimensional measurements of velocity. About the data processing, we use A/D conversion and interface with a microcomputer, so that the data can be processed automatically by the microcomputer. It not only omits some complicated instruments, but also makes the measurement faster and more accurate. The preliminary experiment was performed and the experimental results show the usefulness of the present velocimeter for measurements of the velocity.

This paper presents an industrial monitoring and controlling system designed for the size measuring of the automation line. The system employs a linear CCD image sensor and 8098 single-chip computer. The object is imaged on the CCD through an optical setup, then, the image signal of the object is produced. By means of binarilizing, amplifying and transforming, the pulse signal with certain width corresponding to the length of the object is inputted to the single-chip computer. The system applies the method of filling pulse. By calculating the numbers of the pulses filled, the system obtains the length of the object and displays it on a light-emitted diode (LED). When the size of the object is over the tolerance, the system will warn by audio and light and operate the take-off device to take the object out of the line. The measuring range is adjustable. Therefore, the system is suitable for real time industrial application.

The use of a curved array airborne ultrasound transducer for autonomous vehicle navigation is described. The physical array configuration enables beam steering, beamshaping and beamforming to be efficiently implemented, and this implementation is discussed. A procedure for dynamically evaluating the local environment to determine the most appropriate direction and beam shape is proposed. Simulation results are presented that compare this technique with conventional scanning procedures.

Design of a fuzzy controller requires specification of both membership functions and decision rules. Specification of membership function for a fuzzy logic controller has been an important issue. The traditional way of selecting membership functions has been, in most cases, an ad hoc procedure. In this paper, an optimization algorithm based on simulated annealing for designing a fuzzy membership function for fuzzy controllers is introduced. An optimization algorithm for designing the membership functions and fuzzy rule base for a fuzzy controller is presented. To demonstrate the method, a fuzzy controller for the truck backup problem is designed and implemented using this procedure.

The objective of this work is to generate a path plan for an automatic laser equipped burn debridement robot. A high powered, CO₂ based, pulsed laser transmitted through a kinematic chain to the end effector of a robot is used to ablate burnt tissue on the skin of a burn victim. The generation of the path plan for the robot's end effector is critical for an effective treatment using this automated method. Path planning for this device involves generating the trajectory to be followed by the end effector so that the laser irradiation can be used to treat (ablate) the non-uniform burnt skin without an over- or under-cut. In this application, 3-D body surface points of the human body are measured and modeled using the non-uniform rational B-splines (NURBS) surface modeling technique. The depth of burn is measured along the normal direction to the cutting point on the surface. The homogeneous path is generated under the following constraints: the laser spots cover the burnt area with minimum overlap; the laser beam is incident along the direction normal to the skin; and the laser dosage is varied with the depth of burn. The path traced by the laser treatment head (on the end effector) has a constant offset from the point in the burnt skin being treated. The offset contours on the skin surface are equal in distance. The axis of the laser treatment head is along the normal direction to skin.

The purpose of this work is to employ multiple modular control laws to enable a robot to move through a constrained environment and recognize the objects that are encountered. The two major tasks assigned to the robot are data collection for navigation and object recognition. These basic functions are the tasks involved in accomplishing many robotic applications. Data is collected in the work space by sensing the position of objects of interest. The robot has the capability to sense position in two ways. The robot may touch the end effector to a point in the work space and note the position of the contact point. The robot may also identify objects visually through use of a camera. The advantage of the system presented here is the modular design of the controllers. The system may operate using only one or a combination of the sensors. The decision regarding which sensors are appropriate is made according to the environmental conditions and the task requirements. The system makes this decision by computing an associated numerical measure of the confidence of success and processing cost for each sensor. The system is tested using two different experimental situations to show that the decision criteria applies in general. The system determines how to use the sensors to gather data in an inspection experiment and a maze navigation experiment.

The key technology of future advanced manufacturing systems is the manufacturing cell intelligent quality control (MCIQC). First, this paper focuses on the basic problems such as the overall plan and composition of MCIQC, and then discusses online sensor technology, static measurement of geometrical accuracy and error correction of machine, dynamic error separating and correcting technology, expert system for measuring geometrical parameters, and so on. Finally, the living example of a grinder is used to explain one actual intelligent quality control system.

A CCD based barcode reader has been developed to read the POSTNET barcode with a bar/space width as small as 13 mils. This barcode is currently being used by the united States Post Office. The system can decode barcodes up to a conveyor speed of 250 ft/min for bars travelling parallel to the CCD sensor array. The system consisting of a camera and a logic unit was designed to incorporate various lengths of linear CCD sensor arrays manufactured by EG&G Corporation. The length of the sensor is dependent upon the required field of view. The camera unit processes the analog signal from the CCD sensor and converts it into a binary signal, which is then transmitted to the logic unit. The logic unit uses a Texas Instrument TMS320C30 processor and does the actual signal processing and decoding of the POSTNET code. This paper describes the hardware and the software developed for this system using a 512 element CCD sensor.

An intelligent vibration gripper has been developed. The gripper can handle an object of unknown weight and unknown surface conditions with the most suitable grasping force. Additionally, the gripper can adapt the grasping force to the most suitable value for the object even if the weight and the surface conditions of the object change. The gripper has two parallel fingers. Each finger is composed of a mechanical vibration system with two mass blocks and two springs. In one mass block, a harmonic force is always generated and is absorbed by a dynamic damper consisting of another mass block and a spring, so that one mass block is kept in a stationary condition. The object is grasped by stationary blocks on both sides. When the object is grasped with the most suitable grasping force, the object and all of the mass blocks become stationary because harmonic forces are transmitted to the object and the two harmonic forces cancel out each other. In this gripper, the grasping force is adjusted to the most suitable values by detecting the amplitudes of the dynamic dampers. This paper describes a fundamental principle of a new vibration gripper.

Orientation and stacking of parts has been mostly based on their external features. There are a number of parts which need to be oriented and stacked based on internal features, either because internal features are of primary importance or because of lack of appropriate external features. Issues concerning the orientation and stacking of parts based on their internal features are investigated and two-dimensional axi-symmetrical parts have been classified into groups based on both internal and external features. Fundamental methodologies have been developed for orienting and stacking of parts based on internal features and experimental verification has been performed for many cases. Results indicate that parts can be oriented efficiently and at high speeds. In addition, it can be concluded from these experiments that in many cases, the widely accepted rule that orientational position of parts should be preserved does not necessarily hold.

This paper presents a comparison of three differing methods applied to the analysis of control data from a high speed machinery application. The source data and the pre-processing applied to improve the suitability of the data to the analysis techniques is discussed. The methods compared are cluster analysis, multi-layer perceptron neural networks and self organizing feature maps. The aim of the work is to determine the merits of the techniques in separating normal running operation from faulty operation. The methodology used with each technique is explained and results are computed so as to give the fairest comparison of their respective abilities. Additionally, the ways in which such techniques would be integrated into a final system for the analysis, diagnosis, and control of a high speed machine to give improved reliability are discussed.

A concept of distributed actuators and their controls for ultra-fast, material transfer network based upon linear induction motors (LIMs) is presented in this paper. The distributed actuator consists of modularized stationary primaries connected to form a distributed, closed guideway structure with multiple loops through which vehicles (moving secondaries of the LIM) are propelled at a maximum velocity of 120 km/hr and a maximum acceleration of 4 g. The turning radius can be less than 1 m. A transverse flux LIM propulsion unit design provides robust, direct-drive electromagnetic vector controlled propulsion and electromagnetic steering. The controlled moving secondary vehicles are completely passive with no power, sensing, or control intelligence onboard. The need to control multiple passive vehicles simultaneously moving through a multitude of propulsion units mandates a highly distributed control architecture where propulsion units cooperate using a communication network. The need for centralized database and control is eliminated by endowing the communication network with a topology identical to that of the guideway network. The distributed control architecture ensures spontaneity of material delivery by integrating vehicle motion control with intelligence for real-time collision avoidance, a parallel shortest-path routing algorithm, and an autonomous dispatching protocol. Recent experimental results are presented.

A multi-finger robot which can handle three dimensional objects of unknown weight and unknown surface conditions with the most suitable grasping force has been developed. The hand has three fingers. Each finger is composed of a torsional vibration system consisting of a damping roller, a vibration roller, a rotational support spring, and a rotational absorber spring. A small motor with an eccentric mass mounted on its shaft is placed in each vibration roller. Harmonic torque is generated in each vibration roller by rotating the eccentric mass, so that the vibration rollers and damping rollers are forced to vibrate. Through the initial grasping process, the object is handled by the vibration rollers with the most suitable grasping force. In this paper, the most suitable grasping force is derived through theoretical analysis.