Proceedings of the Institution of Mechanical Engineers,Part I: Journal of Systems and Control Engineering

Increasing demands for higher operational speeds, the need for flexible machinery and recent developments in microchip technology have made programmable machine systems an attractive alternative to conventional systems. However, some difficulties still remain for proper control of programmable systems, especially at higher speeds. These can be categorized into two groups: trajectory planning and trajectory tracking. Conventional trajectory planning methods are ineffective for general application, especially when velocity and acceleration conditions are included. There are many mathematical functions but polynomials are shown to be the most versatile for trajectory planning; however, these can give curves with unexpected oscillations, commonly called meandering. Tracking of a motion in this situation could engender severe practical problems. In this study, a new interpolation method using polynomials with arbitrary powers is proposed to overcome this disadvantage.

This paper addresses the many issues which confront a designer of a process machine when independent servo-drives are to be incorporated. Actuators which meet process requirements and which maximize servo-drive potential are described. Design and selection criteria for all elements of the total transmission system are discussed: servo-motors, feedback devices, control loop strategies, couplings, gearboxes, transmission belts, motion profiles and control computer systems. The paper draws upon a practical system which incorporated ten servo-drives cycling at 13 Hz with positional accuracies of 0.4 mm at 2.1 m/s.

Hydraulic systems are commonly used in industry when large forces of torques are required. Increasing demands for positional, force and speed control from users have led to the use of closed-loop control techniques. While classical controllers are used successfully for many hydraulic applications, they cannot cope with the non-linearities inherent in hydraulic systems or the changes to system parameters over time. This paper considers these problems and proposes a solution in the form of a neural network-based controller in the forward path of the system directly controlling the hydraulic plant. A network-learning mechanism is also proposed to train the network to minimize the system error and examples are given of the implementation with comparisons against a PID controller. The examples illustrate the rapid convergence of the training algorithm and the robustness properties of a simple two-neuron, two-input network controlling a non-linear time-varying plant.

Current state of the art in the design of high-speed machinery for the production or processing of discrete products often involves the use of independent drives synchronized through controllers, rather than via stiff mechanical connections. High-speed machinery for discontinuous processes tends to be characterized by the following attributes: (a) synchronization is highly critical between the axes in certain groups; (b) strong coupling between axes (groups) can be introduced by the work material; (c) the speeds of operation are such that computation is at a premium and just be restricted; (d) individual axes have periodically varying parameters (with additional non-periodic noise); (e) individual axes can become strongly non-linear at high torque (or force) rates; (f) slow and steady trends in the plant parameters are common; and (g) the development of reliable, high-fidelity dynamic models of all machine components for perfect design simulation is impracticable. This paper addresses the issue of how controllers may be specified and designed to provide control solutions for high-speed machinery, which provide the designer with a high degree of confidence that simulated performance may be realized in practice. The form of the solution proposed is an adaptive decentralized control scheme with a recursive identifier to track machine parameter variations. H_∞ design methods are used both to specify the form of the control system and to ensure ongoing robust control of the machinery with minimum sacrifice of performance. Three examples are given(two simulation and one experimental) to demonstrate the benefits of using H_∞ methods, rather than traditional methods, for this type of machinery, and one of these illustrates the effectiveness of adaption for maximizing performance while maintaining stability.

The minimal control synthesis (MCS) algorithm is an adaptive control strategy that requires no prior knowledge of plant dynamic parameters, and yet is guaranteed to provide global asymptotic stability of the closed-loop system. The purpose of this paper is to present MCS as applied to web tension und transport control a class of plant that has highly non-linear dynamics and time-varying parameters. The plant is difficult to control by conventional methods over its full operating range. A typical example and model of such a plant is presented along with the implementation of MCS. Experimental comparisons of MCS with conventional control benchmarks are provided. It will be seen that MCS significantly outperforms the conventional controller.

Switched-reluctance motors appear to be ideal industrial prime movers capable of precision speed and position control. The efficiency can be higher than for a similar-sized induction motor and the electronics less complicated for precise speed control. While the switched-reluctance drive is common in some applications, it has not been widely accepted because of the large amount of torque ripple produced. The torque ripple from the widely used induction motor is quite low and it causes less vibration in the mechanical drive train following the motor. A four-phase switched-reluctance motor can he operated in such a way as to produce a constant zero-ripple torque output. The currents in at least two of the four phases are set so that the total torque produced is constant. By precisely setting the currents in three of the four phases, a constant torque output can be obtained at a constant d.c. supply current, and the switched-reluctance motor then has similar characteristics to a d.c. series motor. A mathematical description of these non-linear currents is derived along with the individual and mutual torque contributions to the total constant torque. The equations are also shown in graphical form.

EFFICIENT DYNAMIC SIMULATION OF ROBOTIC MECHANISMS. Kathryn W Lilly, Kluwer Academic Publishers, Dordrecht, The Netherlands: 1992, pp 160, £45 (ISBN 0 79239 2868)

NEURAL NETWORKS IN ROBOTICS. Edited by George A Bekey and Kenneth Y Goldberg, Kluwer Academic Publishers, Dordrecht: 1993, pp 563 £77.50 (ISBN 0 79239 268 X)