This study investigates the application of adaptive time-suboptimal positional control to an electrical machine with a wide range of loading conditions. These unpredicatable load conditions included variable system parameters, such as inertia variations and nonlinear amplification gain in the servo driver, as well as external disturbances, including viscous frictional force, coulomb frictional force and static loading torque. The design objective was to provide an extremely fast positional movement to the desired target without overshoot and zero steady-state error over these loading conditions. The resultant microcontroller-based adaptive controller consists of an on-line parameter estimator and a robust time-(sub)optimal position controller. The system parameters are estimated by an recursive least squares (RLS) estimator during the acceleration phase. The sampling frequency used by the RLS algorithm is determined adaptively. During the crusing phase of the positional movement, the estimates are further improved by feeding intersample data (stored during the acceleration phase) through an off-line RLS estimator. The coulomb friction and the static loading torque are effectively compensated by a simple mechanism. Another novel mechanism which takes account of nonlinear amplifier gain has also been developed. The time-(sub)optimal position controller calculates the desired reference trajectory in real-time and directs the system state to the reference trajectory. The above adaptive control scheme was implemented on a microcontroller-based system and was applied to an experimental system consisting of a 500W DC permanent magnet motor fed by a pwm servo driver. Experimental results revealed that the proposed controller adapted well to changes in inertia, viscous friction, coulomb friction and amplifier nonlinearity, and the desired time-suboptimal respones were obtained in all these loading conditions.

Many controller tuners are based on linear models of both the controller and process. Desired performance is often predetermined or adjusted in a manner which is not directly related to the desired response. All physical processes contain nonlinearities, commonly of the actuator saturating type, and many controllers contain heuristics for implementation in real systems, such as anti-integral wind up in PID controllers. For different processes a range of closed loop response shapes are desired, often described by features of the response such as rise time, overshoot and settling time. This thesis investigates the possibility of basing controller tuning on closed loop system response data such that desired performance is incorporated directly in terms of familiar time domain features or labels thus eliminating the need for a mathematical process model and repeated tuning reformulations to achieve the desired performance. A controller tuning method named Label Based Neuro-Control is developed and analysed by application to PID controller tuning for a number of process models indicative of real process behaviour. The Method of Inequalities is employed as a comparative controller tuning technique and observations are made concerning its performance. Simulations and numerical investigation indicate that LBNC is a viable technique for the tuning of low order controllers for SISO processes. Tuning is straightforward, flexible and copes well with process parametric changes and performance specification reformulation. The drawbacks are a complicated pretune phase, a limited selection of suitable labels and a difficulty in defining general classes of tuning problems for its application. The method of inequalities is shown to be a powerful technique applicable to higher order controllers and provides a natural incorporation of system constraints. However, operator supervision is necessary for successful tuning. Neither technique is based on the assumption of process linearity but due to the inability to characterise classes of input signals and operating points the types of process nonlinearity are restricted. The controller may be nonlinear, but must be predetermined, and an input/output process model of arbitrary structure is required.

Computer-based systems are being used to tackle increasingly complex problems in ever more demanding domains. The size and amount of knowledge needed by such systems means they are becoming unwieldy and difficult to engineer into reliable, consistent products. One paradigm for overcoming this barrier is to decompose the problem into smaller more manageable components which can communicate and cooperate at the level of sharing processing responsibilities and information. Until recently, research in multi-agent systems has been based on ad hoc models of action and interaction; however, the notion of intentions is beginning to emerge as a prime candidate upon which a sound theory could be based. This research develops a new model of joint intentions as a means of describing the activities of groups of agents working collaboratively. The model stresses the role of intentions in controlling agents� current and future actions; defining preconditions which must be satisfied before joint problem solving can commence and prescribing how individual agents should behave once it has been established. Such a model becomes especially important in dynamic environments in which agents may possess neither complete nor correct beliefs about their world or other agents, have changeable goals and fallible actions and be subject to interruption from external events. The theory has been implemented in a general purpose cooperation framework, called GRATE*, and applied to the real-world problem of electricity transportation management. In this application, individual problem solvers have to take decisions using partial, imprecise information and respond to an ever changing external world. This fertile environment enabled the quantitative benefits of the theory to be assessed and comparisons with other models of collaborative problem solving to be undertaken. These experiments highlighted the high degree of coherence attained by GRATE* problem solving groups, even in the most dynamic and unpredictable application contexts.

This thesis presents a dynamic analysis and a control system for a flexible space manipulator, the Deployable Robotic Manipulator or DRM, which has a deployable/retractable link. The link extends (or retracts) from the containing slewing link of the manipulator to change the DRM's length and hence its workspace. This makes the system dynamics time varying and therefore any control strategy has to adapt to this fact. The aim of the control system developed is to slew the manipulator through a predetermined angle given a maximum angular acceleration, to reduce flexural vibrations of the manipulator and to have a certain degree of robustness, all of this while carrying a payload and while the length of the manipulator is changing. The control system consists of a slewing motor that rotates the manipulator using the open-loop assumed torque method and two reaction wheel actuators, one at the base and one at the tip of the manipulator, which are driven by a closed-loop damping control law. Two closed-loop control laws are developed, a linear control law and a Lyapunov based control law. The linear control law is based on collocated output feedback. The Lyapunov control law is developed for each of the actuators using Lyapunov stability theory to produce vibration control that can achieve the objectives stated above for different payloads, while the manipulator is rotating and deploying or retracting. The response of the system is investigated by computer simulation for two-dimensional vibrations of the deployable manipulator. Both the linear and Lyapunov based feedback control laws are found to eliminate vibrations for a range of payloads, and to increase the robustness of the slewing mechanism to deal with uncertain payload characteristics.

This thesis investigates a number of problems of the implicit linear systems framework. First, the problem of realisations of nonproper transfer functions is considered. The main result obtained here is the generalisation of the realisation method from MFDs to the case of the nonproper transfer functions. The obtained realisations are singular systems. The method treats both finite and infinite frequency behaviour in a unified way and generalises the results related to the minimality of the realisation and coprimeness and column reducedness of the MFD. Furthermore, it displays transparently the relation between the extended MacMillan degree of the transfer function and the minimal realisation. The next problem considered is the problem of canonical forms of minimal singular systems under restricted system equivalence transformations. For systems with outputs a canonical form is obtained and it is shown that it is directly related to the echelon form of the composite matrix of an MFD of the transfer function of the system. This result is a direct generalisation of the results of Popov and Forney for strictly proper systems. The canonical form obtained is of Popov type and may be considered as a direct generalisation of the well known form for strictly proper systems. The second canonical form is for systems without outputs. A Popov type canonical form for a class of these systems is obtained. This class is that of systems with equal reachability indices. For both canonical forms, the sequence of the transformations yielding the canonical description is described in detail. In the general case of systems without outputs a semi canonical Popov type form is obtained. Another problem considered in the thesis is the problem of first order realisations of autoregressive equations within the external equivalence framework. An alternative to the existing methods is provided; in fact, the proposed method is simpler than the existing ones and allows the derivation of the realisation by inspection of the autoregressive equations. A generalisation of the observability indices is proposed for nonsquare descriptor systems and their connection to the autoregressive equations is established. The problem of model matching for implicit systems is considered next. This is a generalisation of the model matching problem for systems described by transfer functions. Here a controller is interconnected to the given plant such that the overall system has a desired external behaviour. The problem is studied within the framework of external and A-external (input-output) equivalence. Necessary as well as sufficient conditions for the solvability of the problem are derived and the equations of the controllers are found in a constructive way. The last problem considered here is the generalised dynamic cover problem of geometric theory i.e. the problem of finding the family of (A, B)-invariant subspaces covering a given subspace. This problem is formulated here by using the matrix pencil approach of the geometric theory. This approach allows the unification of the problem for state-space and nonsquare descriptor systems. An extension of the problem to the case of infinite spectrum spaces is also obtained. The solution of the above problems is reduced to the solution of appropriately defined systems of linear equations. Finally, an alternative method for the solution involving systems of multilinear equations is proposed using the mathematical tool of Groebner bases.

Inverse simulation is a technique by which the necessary control actions can be calculated for a vehicle to perform a particular manoeuvre. The manoeuvre definition is thus the input to the problem, and the output is a time history of the control motions. The realism of a result is clearly dependent on the fidelity and sophistication of the vehicle mathematical model. Present inverse simulation algorithms are limited by being model specific and only able to accommodate models of restricted complexity. For helicopters specifically the models used in inverse simulation are, in general, rudimentary in nature. The existing inverse simulation algorithm at Glasgow University, "Helinv" is specific to the helicopter model, "HGS". Though HGS is very advanced by comparison with other inverse simulation helicopter models, it lags far behind the state of the art in conventional simulation. The principal aims of this research were therefore twofold: to develop a robust, generic inverse simulation algorithm, "Genisa"; and to develop a state of the art individual blade helicopter rotor model, "Hibrom". Naturally verification and validation were integral to these aims. These objectives having been achieved the intention was to demonstrate the flexibility of Genisa and the value of Hibrom by performing inverse simulations of various rotorcraft configurations. As well as representing a novel tool in rotorcraft simulation, the development of a flexible inverse simulation algorithm which can accommodate complex models extends the boundaries of inverse problems in general. Genisa has proven to be both flexible and robust. Hibrom has been verified, validated and - using Genisa - successfully used in inverse simulation. The advantages of an individual blade model in inverse simulation have been demonstrated by comparing results with the disc model, HGS. Inverse simulations have been performed for various rotorcraft configurations identifying the respective benefits of the different vehicles. In all respects the aims identified above have been met in full.

Correlation techniques for the identification of nonlinear systems are discussed in Chapter 1. The Volterra series expansion of the response of a nonlinear system is described, together with its counterpart in the frequency domain. Qosscorrelation methods for identifying the kernel functions which occur in this expansion are reviewed, with particular emphasis on techniques or obtaining the linear approximant to a nonlinear system. A crosscorrelation method is also discussed which appears to be unrelated to the Volterra approach. This technique uses a 4-level test signal and is the subject of detailed analysis in later chapters. The 4-level test signal is discussed in detail in Chapter 2 and it is shown that the linear channel of a nonlinear system may be identified. The concept of the almost periodic function is presented and it is concluded that by means of almost periodic fi.ncti&ns it is possible to identify the linear portion of the nonlinear channel in the system. The technique is essentially based on destroying the ynchronisation of the two 2-level signals originally used to produce the 4-level signal. A generalised technique is developed to calculate the parameters of the 4-level aperiodic signal in order to apply the method to any single-valued nonlinearity, assuming that 'a priori' knowledge of the nonlinearity is available. The technique is extended to permit identification of the impulse response of the linear elements when the nonlinearity is single valued but contains only even components. This is achieved by modifying the nonlinear characteristic to provide an odd component. A further extension of the technique shows that identification of certain nonlinear channels containing elements with memory is possible. Several situations are analysed in detail. An extension is also considered where the nonlinear channel is composed of two linear transfer functions separated by a nonlinearity. Experimental results are included to show the accuracy of the techniques developed. In Chapter 4 the application of high frequency signals (dither) as a technique to be used in the identification of certain open and pl,osed loop nonlinear systems is discussed. It is shown that certin types of dither permit linearization or elimination of nonlinear channels in certain open and closed systems. It is therefore possible, under these conditions, to identify the linear portions by impulse response or frequency testing techniques, and experimental results show the accuracy achieved. The concept of the equivalent nonlinearity provides a simple interpretation of the action of the dither. Dither is next considered in its role as a means of stabilizing nonlinear control systems, and the particular case of a third order system containing a hysteresis type relay characteristic is analysed in detail. The concepts of equivalent nonlinearity and describing function are used to derive an expression for the minimum amplitude of dither required to quench limit cycle oscillations in the system. The system is shown to hAve an effective gain margin once oscillations have been quenched. A typical point in this region is investigated and the three dimensional domain of stability for the system is presented,together with typical trajectories. The system is also shown to exhibit subharmonic resonance apci jump phenomena. In chapter 6 the use of dither in adaptive control systems is discussed, and the specific case of dither adaptive control system proposed as the solution to a qonstant fuel-rate problem is analysed in detail. The dynamic analysis presented divides the system into two loops via an extension of the equivalent nonlinearity concept. The final chapter reviews the work described in the body of the thesis, and explores several avenues for future research.

Part 1: Synergic control in MIG welding. Synergic controls (i.e. real time control methods) for both steady DC open arc and short circuiting arc operations have been developed in this study. These have been based on the generalised quadratic and linear ‘power-current' and linear ‘voltage-current’ equations which adequately describe the MIG welding operation. The controls are low coat electronic units which are added to a ‘transistor’ controlled power source. The units based on the ‘power-current’ concept can operate with the power source set in the constant current, constant voltage or any intermediate mode of output characteristics, and regulate only the steady DC open arc operation. The unit based on the ‘voltage-current’ concept operates with the power source set in nominally constant voltage mode, and regulates both steady DC open arc and short circuiting arc operations. These controls adjust the current, voltage or power automatically according to any operator-selected wire feed speed whether maintained constant at any level, varied gradually or modulated with any waveform to achieve ‘thermal pulsing’. In addition, two approaches have been used to adapt the synergic control units to a lower cost, thyristor controlled power source which is more widely used in industry. In one approach, the generalised control equation used previously for the transistor controlled power source has been extended to take account of the output characteristics of the thyristor controlled power source. As an alternative, the control units themselves have been modified to allow for the different power source characteristics. The control systems have been successfully demonstrated for producing sound welds in a wide range of welding applications, for both mechanised and manual welding techniques. Part II: Penetration control in TIG welding A 'backface' penetration control system developed at Liverpool University, based on a ‘video camera' instead of a ‘photodiode' as a sensor, has been evaluated at The Welding Institute. Essentially, the system controls the size of the weld pool, instead of an average level of radiation, by regulating pulse current period, by means of a fibre optics image guide/video camera/microcomputer based controller. The system has been found to be capable of controlling the weld bead penetration uniformly in stainless steel plates and joints, for constant as well as variable material thicknesses. However, the system cannot be used with high frequency (HF) arc discharge, as required for automatic arc initiation, because the software and electronic components are corrupted. A hard-wired control unit has therefore been developed at The Welding Institute to replace foe microcomputer based controller. This unit can be operated reliably with automatic HF arc initiation. Furthermore, it has been shown to control weld bead penetration in stainless steel plates and joints, for constant at well as variable material thicknesses.

Control System Design methods are presented in terms of optimization techniques that incorporate Multi-Objective design criteria. Computer-Aided Control System Design (CACSD) environments make the approach easy-to-use and accessible to the practising control engineer. Two CACSD environments have been developed using different versions of the MATLAB package, one interfacing the ADS optimization package to an upgraded FORTRAN version of MATLAB, the other using Non-linear Programming algorithms coded in the PRO-MATLAB command language. In both environments, optimization problems are entered interactively and in a flexible manner using simple interpreted commands and programs. A Control System Design method has been implemented using optimal control theory and integral quadratic measures of control. The theory has been developed to incorporate a large number of design options, control structures and disturbance types. An evolutionary design process is used so that the control order and number of design criteria are systematically increased to incorporate more complex control structures and a wide set of performance objectives. In the later stages of this evolutionary design process, a Multi-Objective design strategy, known as the Goal Attainment method, is used to address multiple performance objectives.

No description available.