Annual Review in Automatic Programming最新文献

In recent years fuzzy logic is increasingly applied to the lower levels of process automation. To identify applications in which fuzzy systems are useful in contrast to other knowledge based methods, it is necessary to distinguish the advantages from using expert knowledge to specify a system behavior in contrast to a formal description and those which arise from the fuzzy interpretation of knowledge. This paper tries to examine this question under aspects like control results, ease of handiness, and implementation. It uses the well known example of the pole balancer problem.

On-line process fault diagnosis using fuzzy neural networks is described in this paper. The fuzzy neural network is obtained by adding a fuzzification layer to a conventional feed forward neural network. The fuzzification layer converts increments in on-line measurements and controller outputs into three fuzzy sets: “increase”, “steady”, and “decrease”. Abnormalities in a process are represented by qualitative increments in on-line measurements and controller outputs. These are classified into various categories by the network. By representing abnormalities in qualitative form, training data can be condensed. The fuzzy approach ensures smooth transitions from one fuzzy sets to another and, hence, robustness to measurement noise is enhanced. The technique has been successfully applied to a CSTR system.

This paper proposes a performance design method of real time system with use of three variation of Bottleneck Diagnosis Expert Systems (BDES), which generate queueing network models of target systems, identify the network bottlenecks, and suggest the designer how to improve the performance of the networks, with use of qualitative reasoning technique. After input of network topology and parameters, these tools show the designer the bottleneck and how to improve it automatically.

RH&H's role in the HINT project was to focus on integration, finishing, utilisation, and marketability of the overall technological result. The other HINT partners were responsible for separate AI techniques and system units, which RH&H integrated to create a commercial product. To identify the industrial requirements of a commercial product, RH&H carried out a survey based on qualitative interviews with experienced, technical managers. We realised that to turn a R&D Project into a commercial product, it is crucial first of all to understand the industrial functional needs and the industry's way to conceptualise the world.

This paper presents the integration of Artificial Intelligence representation techniques and high level Petri nets in a knowledge based manufacturing modeling and analysis system. By means of Petri nets and its implementation based on rules, a model of the system can be constructed. In order to verify the real behavior, the control system interchanges signals with the plant. When an abnormal situation is detected, the control system must take control decisions to allow the system to work in the new situation. Error management in these cases includes detection, diagnosis and reaction. In this paper the hierarchy of object classes to control the system is enlarged to face with the problems produced by abnormal behavior.

This paper presents differents ways of using artificial neural networks in adaptive control. A classification of architectures for control using neural networks is presented, showing the existing paralelism with Adaptive Control techniques.

In this paper, Fuzzy Control Systems with consecuent functions are analized using concepts from qualitative theory of nonlinear dynamic systems. Particularly, linear weigthed output fuzzy controllers and weight-output fuzzy controller are considered.The consequent function coefficients of rules around the equilibrium point are related with the stability and robustness indices. Finally, a explanatory example is presented.

Implementation of process control of pulsed laser deposition (PLD) requires actuators, sensors and a “materials and processing” knowledge base. Improvements in quality and reproducibility of material deposits are achieved by using this knowledge base to drive the process toward desired operating regions. The knowledge base includes experimentally determined, empirical relationships describing the complex dynamical interactions of laser parameters. Intelligent control based on this description can be implemented to reduce process disorder and efficiently produce consistent coatings with desired properties.

Our research deals with the intelligent supervision and control of batch fermentation processes. We built a pilot plant, with all the auxiliar electronics needed, for a MS-DOS computer to interact with the plant. We selected KAPPA-PC as the main platform to develop a real-time application. It is an object-oriented development environment, with the tools to define and manage class hierarchies including sets of rules, for intelligent functions. The shell works under MS-Windows, and is open to userdeveloped specific C code. After some experimental modelling, we determined models, and defined intelligent classes to simulate the plant, to control the process, and to supervise it, in real-time.

An approach for designing the artificial intelligence part into the human-machine interface is described. Two application domains, namely chemical and cement plants, are considered. Interactive techniques for a better involvement of the end-user in earlier design phases are described. They include task and knowledge analyses, participative rapid prototyping design and usability testing. The logical design of knowledge based functionalities of the human-machine interface is exemplified with tracking interaction and supervising dialogue. The support tools for participative design are also briefly mentioned.