Many approaches have been developed for estimating those variables which are difficult to measure online in industrial process situations. In this paper, two approaches that can be used to provide frequent and accurat...
详细信息
Many approaches have been developed for estimating those variables which are difficult to measure online in industrial process situations. In this paper, two approaches that can be used to provide frequent and accurate estimates of process outputs which are subject to large measurement delays are outlined. The first is based upon linear adaptive techniques whilst the other makes use of a fixed parameter neural network model. The development and application of the two estimators is addressed. The results from recent industrial application studies and plant simulation studies serve to highlight the characteristics of the different philosophies taken in estimator design. Moreover, this allows a comparison of the performance capabilities of the two techniques.< >
Although optimal control techniques provide, in theory, the ability to design suitable control strategies for general control problems, in practice the results are not commonly implemented due to the large computation...
详细信息
Although optimal control techniques provide, in theory, the ability to design suitable control strategies for general control problems, in practice the results are not commonly implemented due to the large computational requirements for solving such problems. A recent development in computer systems has been the use of parallelprocessing techniques to speed up computations by sharing the work load over several processors that can interact between each other for message passing, sharing variables etc. In particular the advent of the INMOS transputer with its unique serial links have opened the doors to true parallelism in computer systems and much applications research is currently underway that is aimed at using transputer arrays. An area of interest to the authors is automatic flight control, which could also benefit from the introduction of parallelprocessing. In the paper the authors present two parallel optimal control algorithms for aircraft systems.< >
BNN laboratories is developing the concept of a real-time data acquisition, analysis, and advanced control system based on the Butterfly parallel processor. The Butterfly provides a flexible and powerful vehicle for i...
详细信息
BNN laboratories is developing the concept of a real-time data acquisition, analysis, and advanced control system based on the Butterfly parallel processor. The Butterfly provides a flexible and powerful vehicle for implementation of these concepts. The Butterfly is a shared memory parallel processor which uses unique switch technology.< >
Self-organising structures, considered suitable for the exploitation of AI-related heuristics search strategies are described. Possibilities for implementation of SOC (self-organising control) structures on transputer...
详细信息
Self-organising structures, considered suitable for the exploitation of AI-related heuristics search strategies are described. Possibilities for implementation of SOC (self-organising control) structures on transputers are also investigated, particularly the fuzzy control algorithm, where the merging of an alien language with occam is considered along with the use of LISP as an AI language.< >
Describes the development of techniques using parallelprocessing for the tracking of multiple targets. Some initial studies using transputers on a specific scenario, namely the tracking of two crossing targets, are d...
详细信息
Describes the development of techniques using parallelprocessing for the tracking of multiple targets. Some initial studies using transputers on a specific scenario, namely the tracking of two crossing targets, are described and results given. This problem serves to introduce the procedures being used and which are currently being applied to more complex scenarios.< >
The paper is concerned with the application of parallelprocessing techniques to the eddy current non-destructive testing (NDT) instrumentation based on Kalman filtering and a multiple model estimation Algorithm (MMEA...
详细信息
The paper is concerned with the application of parallelprocessing techniques to the eddy current non-destructive testing (NDT) instrumentation based on Kalman filtering and a multiple model estimation Algorithm (MMEA), which is utilised as a procedure to diagnose and classify defects of different size and depth in a tubular metal sample. parallel decomposition of the algorithm is considered for real time generation by an array of VLSI Transputer processors.< >
Least squares solutions of linear equation occur in adaptive antennas, frequency filters and Kalman filters and Kalman filters. There is now the possibility of economic parallelprocessing to solve such equations and ...
详细信息
Least squares solutions of linear equation occur in adaptive antennas, frequency filters and Kalman filters and Kalman filters. There is now the possibility of economic parallelprocessing to solve such equations and in this paper the application of systolic arrays to these areas is discussed.< >
parallel- and pipeline-processor structures can maximise the speed-up factor and minimise the input/output cost in a network of transputers. These structures can implement most digital signal-processing applications. ...
详细信息
parallel- and pipeline-processor structures can maximise the speed-up factor and minimise the input/output cost in a network of transputers. These structures can implement most digital signal-processing applications. However, the paper concentrates on the evaluation of the discrete Fourier transform (DFT). The number of arithmetic operations and interprocessor transfers are used to evaluate the algorithms. Speed is shown to depend upon the distribution and sizes of data points and upon the numbers of transputers available.< >
A controller for end-effector forces must be powerful enough to handle hybrid position/force control algorithms and inverse dynamics, as well as a six axis multi-variable inner-loop controller. It must have the capabi...
详细信息
A controller for end-effector forces must be powerful enough to handle hybrid position/force control algorithms and inverse dynamics, as well as a six axis multi-variable inner-loop controller. It must have the capability to acquire data from up to twenty-one sensors and output to six actuators. It must allow flexible inter-process communications as well as providing a simple interface between the various levels of control. A fast software development environment is desirable and floating point arithmetic greatly eases data overflow and scaling problems. The author has developed a transputer based real-time controller and measurements show that it should fulfil the computational requirements for force control of an industrial robot. A pipeline architecture has been chosen for data distribution, and show to introduce a negligible delay compared to the sampling period. Reliance on commercially available analogue interface hardware has proved to be a problem, but simple boards can be built to overcome this limitation.< >
The problem of controlling flexible structures is important but challenging due to the infinite number of modes present. For example to suppress vibrations of aircraft fuselage and wings real time processing power bey...
详细信息
The problem of controlling flexible structures is important but challenging due to the infinite number of modes present. For example to suppress vibrations of aircraft fuselage and wings real time processing power beyond sequential computers is required. Data needs to be measured, processed and corrective control action determined based on some mathematical model. This must be repeated at a suitable sample rate. To complicate matters further, the models relied upon can be in error due to parameter variation, modelling errors, etc. It is hence usually necessary to model the structure on-line and perform detailed simulations before control can be attempted. It therefore seems that classical computing techniques are inappropriate for dealing with control problems of this magnitude. In the paper the authors propose an alternative strategy whereby the potential of parallelprocessing networks is exploited. The authors concentrate on the simulation side of the problem and for convenience restrict attention to a cantilever systems since this covers several control applications. Such structures are governed by fourth order PDEs that need to be numerically solved for control purposes. The authors present a new parallelprocessing technique for solving such PDEs on transputer networks.< >
暂无评论