We had a contract from Navy in 1997 to work on the control of piezoelectric actuators. The controller consists of two loops: feedforward and feedback. The feedforward loop consists of a Fuzzy CMAC controller, which is used to compensate hysteresis nonlinearity. Fuzzy CMAC is a new type of neural net developed by Intelligent Automation Incorporated. It has a learning speed that is an order of magnitude faster than conventional multilayer perceptron neural nets. The advantage of feedforward control is that it can increase the system response speed without interfering with the system stability. We used a PID controller in the feedback loop because the feedforward compensation may have some residual errors and having a PID controller in the loop will help to reduce the error even further. In our experiment, we operate an actuator manufactured by Burleigh Instruments in the region of 100 Hz whereas the actuator resonance peak is about 1 kHz. Experimental results showed our approach can achieve excellent linearity.
Hazardous waste remediation often requires the application of new technologies. In recent years, these technologies have undergone rapid development. Intensely scrutinized in terms of design, new technologies surprisingly receive little attention when assessing the hazards they might pose to worker health and safety. A readily available and user friendly safety and health assessment tool for design evaluation can help assure the safety for operators and the public. This concept involves obtaining data for defining a technology, identifying and defining the elements and the hazards associated with these technology elements in terms of sources and types, evaluating the process to identify potential for human errors, building queries to suggest for removing or controlling the hazards, evaluating the interaction of technology elements, and finally, providing a summary of the hazards identified and suggested corrective measures. Thus far, the system was prototyped to a small existing technology of very limited scope. It showed that design-for-safety can indeed be possible using computer systems that are linked to the Internet. The system was researched for its robustness, design integrity, and the advantage of using expert systems for `What-If' scenarios.
The Lister Hill National Center for Biomedical Communications, an R&D division of the National Library of Medicine, has developed a PC-based system for semi-automated entry of journal citation data into MEDLINETM. The system, called MARS for Medical Article Records System, includes many automated features but requires a few manual tasks such as scanning and the entry of certain data that are not located on the scanned page. Now that considerable computing power and speed are routinely available on desktop PCs, we think it may be possible to include speech recognition as an optional user interface to reduce operator burden and to improve speed and quality for document scanning and data entry. We undertook a study to determine if speech recognition was sufficiently accurate, reliable and immune to noise to warrant integration with MARS workstations.
The rapid development of the magazine industry in Sweden has stressed the importance of modular thinking. However, using modular thinking in the magazine architecture is not yet established in the magazine industry despite the need for shorter lead-time and improved content quality. A magazine production can be divided into two phases; the editorial phase and the manufacturing phase. Surprisingly the two different phases are seldom synchronized even if they co- operate on the same product. The aim of this research was to evaluate the use of modular thinking in the product and focus on its manufacturability as well as its influence on relationships between the corporate cultures of the two phases. The company culture is often mirrored in the product and in the manufacturing process. In this research we further develop the results from two previous projects where it was concluded that the magazine's product architecture regulates the work flow speed and that the company's culture is a dynamic factor which both affects and is affected by the product. The theoretical analysis shows that a magazine can successfully be re-designed, integrating both manufacturability and layout in the product architecture. The degree of modularity in the product architecture determines the performance of three parameters: lead-time, quality of the content and the level of relationships between the two corporate cultures.
Here introduces a concept of Holonically Object Oriented System. 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 or getting a help in a sense of analyzing a system, etc. For having these things accomplished, we have tried to understand the target system to be controlled clearly, accurately, and precisely. After having got these information, it is ready to control for many purposes. But usually this method gives us further complexed problems, more time consuming because of the size of a system, gives us comparatively lower robustness. These can be caused by a 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 Holonically Object Oriented System which consists of a holarchy which behaves as a whole as well as a subordinate. And this behavior helps to interpret a huge organization into a system which is flexible, raising a production rate, and high adaptability to changes.
Since almost all of the scheduling problems are NP-hard-- cannot be solved in polynomial time--those companies that need a realistic scheduling system face serious limitations of available methods for finding an optimal schedule, especially if the given environment requires adaptation to dynamic variations. Exact methods do find an optimal schedule, but the size of the problem they can solve is very limited, excluding this way the required scalability. The solution presented in this paper is a simple, multi-pass heuristic method, which aims to avoid the limitations of other well-known formulations. Even though the dispatching rules are fast and provide near-optimal solutions in most cases, they are severely limited in efficiency--especially in case the schedule builder satisfies a significant number of constraints. That is the main motivation for adding a simplified genetic algorithm to the dispatching rules, which--due to its stochastic nature--belongs to heuristic, too. The scheduling problem is of a middle size Finnish factory, throughout the investigations their up-to-date manufacturing data has been used for the sake of realistic calculations.
This paper introduces the different problems related to the choice of rapid prototyping processes, in order to produce some particular parts or tools. Very often, people have no time to test different solutions or they have not experts for a comparative evaluation of the different possible industrial technologies. Of course, one could imagine that it is very difficult for somebody to know all the present technologies at a time, and to evaluate a multi-criteria choice. Another important aspect is that tests are time- consuming and money consuming. The knowledge-based system presented in this paper will propose, from the specifications (numerical model quality, delay, cost, material, quantity, etc...), different alternatives of rapid prototyping processes, which will be discriminated and optimized when considering a combination of the different criteria of the specifications. The proposed system is based on actual machines and technologies, which are really accessible in the company or outside (subcontractors or laboratories), for the realization of each step of the complete processes. In the following paragraphs, the chosen software solutions for knowledge modeling and process choice will also be presented.
In this paper, we identify fundamental issues and challenges in developing a component-based integrated engineering design approach. We classify issues under specification, verification, design, and reusability categories. We identify the properties of the necessary specification models such as genericity, formality, and consistency. Verification issues include integration constraints and other functional and performance aspects. Design issues include architectures of heterogeneous components, hardware/software codesign, and design metrics. Finally, we identify reusability issues as related to building manufacturing systems from reusable, highly generic, and highly parameterized components. We then describe our research approach to address these issues. By establishing a foundation for an integrated system design approach, we consequently improve our ability to specify, model, and verify effective intelligent manufacturable systems as well as develop tools to support integrated system designs.
One critical aspect in electronic packaging is the fatigue/creep-induced failure in solder interconnections, which is found to be highly dependent on the shape of solder joints. Thus predicting and analyzing the solder joint shape is warranted. In this paper, an automatic computer-aided system is developed to simulate the formation of solder joint and analyze the influence of the different process parameters on the solder joint shape. The developed system is capable of visually designing the process parameters and calculating the solder joint shape automatically without any intervention from the user. The automation achieved will enable fast shape estimation with the variation of process parameters without time consuming experiments, and the simulating system provides the design and manufacturing engineers an efficient software tools to design soldering process in design environment. Moreover, a program developed from the system can serve as the preprocessor for subsequent finite element joint analysis program.
`Plasmo', a novel designed fast signal processor based device has been developed. We use this modular system for the immediate evaluation of the quality of laser material processing. With this tool the machining process can be analyzed in real-time. An adaptable algorithm ensures that this system automatically adapts to parameter fluctuations during welding or cutting. A summary of our experience in the use in car industry is presented in this paper. We also present new ideas and first results to expand this system to detect the full spectrum of emitted light during process instead of two wavelengths used up to now.
Weld penetration and fine formation are the major factors for consideration of weld bead quality. The geometry of weld pool contains abundant information about the weld penetration. The weld pool surface is depressed during full- penetration because of arc impulse, and the weld pool surface may be convex during part-penetration or welding with filler. In this paper, we present the surface height and shape parameters for describing the three dimension of weld pool. During pulsed GTAW process, the weld pool image can be obtained through visual sensing system by the illumination of arc light on weld pool. The inverted image of tungsten tip and arc shape can be seen clearly from the weld pool image. The position of inverted tungsten tip varies with the surface height according to the principle of specula reflection. The point of tungsten tip is located to calculate the surface height. The shape of weld pool has been characterized with size and shape parameters, such as pool width, length and a series of rear angles, etc. A simple nonlinear formula with only four parameters is proposed for describing the pool shape, and the regression results are shown with high accuracy. Based on the surface height and geometry parameters of weld pool, the shape of weld pool can be strictly defined, which lays the foundation for further study on process model and weld penetration control.
This paper reports research in compensating for position inaccuracy and flexibility problems in a loosely coupled robot arm by means of machine learning methods. Error sources in the system are studied and problems are described. A number of methods for eliminating problems due to inaccuracy in 2D-space have been previously reported. These methods have been extended to address the problem in 3 dimensions. Utilizing a real time monitoring system, the end-effector position is sensed. The collected data is converted into appropriate error maps. Using a novel machine learning method, the error maps are used to predict system errors and compensate for them. The machine learning engine generalizes the data for the points between the sampled points. The experimental results are presented.
Most of existing 3D-CAD and 3D-CG softwares have problems in intuitive manipulation and representation of 3D curved surfaces. This paper discusses design issues and a prototype systems of an immersive modeler which enables a user to design intuitively and interactively 3D solid objects with curved surfaces by one's hand a finger. Shape deformation is expressed by simple formulas without complex calculation because of skeletal implicit surfaces employed to represent smooth free-form surfaces. A polygonization algorithm that generates polygonal representation from implicit surfaces is developed to reduce the time required for rendering curved surfaces, since conventional graphics hardware is optimized for displaying polygons. The prototype system has shown that a user can design 3D solid objects composed of curved surfaces in a short time by deforming objects intuitively by one's hand and finger in real-time.
To better integrate engineering design and analysis, the multi-representation architecture (MRA) and related methodology have been developed to represent the information transformations between CAD and CAE models. The MRA consists of four representations for increased modularity and flexibility. As one of the representations, solution method models (SMMs) are object-oriented wrappings of tool-specific inputs and outputs that enable highly automated operation of general purpose analysis tools. Outer contexts in the MRA create SMMs from product data and map SMM results back into product-specific terms.
Intensity division interferometers have been long used in displacement measurement with resolution down to 0.01 micron. However, commercial multi-frequency laser interferometers are very expensive and difficult to be embedded in motion control experiments. In this paper, a single frequency Michelson interferometer combined with an 128 X 1 intensity integration photodiodes to produce a high resolution and wide range displacement detection is presented. With the digital signal microprocessor, plus signal filtering technique, the displacement changes to wavelength/2 has been measured by using PZT stack with high voltage driver. With the He-Ne laser as the light source, the precision is about 0.3 micron, the change pattern of the peak maximum of the fringes corresponding to the change pattern of the PZT driving signal, and the best results are obtained with sine-wave at different frequencies. This system has the features of high resolution, broadband, non- contact measurement. Because of lower cost, easier implementation, faster DSP processing, it is very likely to be embedded in the concurrent control system.
The detection of structural defects in a ball bearing using an embedded piezoceramic load sensor and the discrete wavelet transform is presented. A model to predict the output of the embedded sensor was developed and experimentally verified. A mother wavelet was developed specifically for analyzing the response of the bearing and embedded load sensor. This wavelet was found to produce a more descriptive decomposition of the sensor signal than a standard Daubechies wavelet. Furthermore, the relationship between a bearing misalignment and the resulting load variation was established.
Today's situation with increasingly shorter time-to-market limits and growing variant spectra calls for advanced methods in the manufacturing domain. A big potential for gaining faster and better manufacturing results lies in the application of offline programming, especially if processing small lot sizes. Offline programming offers as main advantage a notable reduction of deadlock times of manufacturing systems. Applying this technology there is no time consumptive teach-in on the robots necessary. A technology module based on CAD/CAM technique--mainly for 3D welding applications--is described which permits to carry out offline path and process planning including simulation and visualization of the processing task.
We have demonstrated a color recognition system composed of a novel, self-organized optical neural network system that includes genetic algorithms along with back propagation schemes, which make it possible that the system avoids the local minimum problems and make the learning processes faster and better. Our system is composed of a color liquid crystal display panel, bistable semiconductor lasers, photo- diodes, and liquid crystal light projector.
The magnetic-induction method of quality control of seamless pipes in real-time characterized by a high level of structural noises having the composite law of an elementary probability law varying from batch to a batch, of a varying form. The traditional method of a detection of defects of pipes is depend to usage of ethanol defects. However shape of actual defects is casual, that does not allow to use methods of an optimum filtration for their detection. Usage of adaptive variants of a Kalman filter not ensures the solutions of a problem of a detection because of poor velocity of adaptation and small relation a signal/the correlated noise. For the solution of a problem was used structural Adaptive Neuro-Fuzzy Inference System (ANFIS) which was trained by delivery of every possible variants of signals without defects of sites of pipes filed by transducer system. As an analyzable signal the error signal of the prognosis ANFIS was considered. The carried out experiments have shown, that the method allows to ooze a signal of casual extended defects even in situations when a signal-noise ratio was less unity and the traditional amplitudes methods of selection of signals of defects did not determine.
Wire-cut EDM is a highly complex process, which is characterized by non-linear behavior. Due to very large number of input parameters, a Neural Network Approach is followed in this paper to model this process. A large number of experiments have been carried out to train the network. The model is also tested using additional experimental data. It is found that neural network models of wire-cut EDM process provide results with required accuracy. The important input parameters have been identified and reported in this paper.
In today's global economy, there is a significant paradigm shift to collaborative engineering design environments. One of key issues in the collaborative setting is the consistency model, which governs how to coordinate the activities of collaborators to ensure that they do not make inconsistent changes or updates to the shared objects. In this paper, we present a new consistency model which requires that all update operations will be executed in the casual order (causality) and all participants have the same view on the operations on the shared objects (view synchrony). A simple multicast-based protocol to implement the consistency model is presented. By employing vector time and token mechanisms, the protocol brings the shared objects from one consistent state to another, thus providing collaborators with a consistent view of the shared objects. A CORBA-based on-going prototyping implementation is outlined. Some of the related work are also discussed.