Department of Mechanical Engineering

Control and actuation Phd theses

Fluid metering using active materials

Dhinesh Sangiah, 2011


Servovalves are compact, accurate, fast flow modulating valves widely used in aerospace, defence, industrial and marine applications. However, cost reduction pressures exist due to tight tolerances required, particularly in the first stage of the valve. In this research novel servovalve concepts are investigated. In particular, a new first stage actuator assembly is developed to move a servovalve spool using the jet principle.

The conventional torque motor assembly in the first stage was replaced by a multilayered bimorph actuator. A feedback wire was used to facilitate proportional flow control via mechanical feedback. The bimorph was directly coupled to the feedback wire for submerged operation. A steady state analytical model of the bimorph-feedback wire assembly was developed to derive the stiffness constants influencing the deflector and the valve spool. The derived stiffness constants were compared to FEA predictions. The flow forces acting on the deflector were determined using CFD analysis. The flow force was found to be proportional to the pressure drop across the deflector and the deflector displacement.

A high order nonlinear model of the valve was developed and used to simulate valve dynamic characteristics. The high order model was linearised and reduced to a first order lag to identify the system parameters that determined the first break frequency and the steady state gain of the valve. Two 'Mark-1' prototypes were built and tested. The measured frequency responses of the prototypes were in good agreement with the simulation results. At 140bar supply pressure and maximum applied voltage amplitude the -3dB bandwidth of the valve was measured at approximately 41Hz. The frequency response of the valve spool was reasonably consistent for varying applied voltage amplitudes at a fixed supply pressure. The hysteresis of the second stage spool was approximately ±4%. The stroke of the second stage spool was approximately 0.42mm.

The bandwidth and steady state gain of the valve were expressed in terms of ratios between the forward and feedback path variables of the valve system. Performance plots were developed using these variable ratios. A 'Mark-2' prototype valve was developed and tested, intended to possess a higher bandwidth. The -3dB bandwidth of the 'Mark-2' valve spool at 140bar supply pressure and maximum applied voltage amplitude was measured at approximately 60Hz. At this voltage amplitude the spool stroke was approximately 0.24mm and the valve hysteresis was approximately ±2%. Bandwidth and stroke were consistent with the predictions.


The optimisation of a high-speed servomechanism

Robert Rayner, 2010


The problem of mechanism design is one of the fundamental, classical engineering problems. The ability to create mechanical devices capable of performing specific tasks and to follow desired motion paths, lies at the heart of the discipline. Numerous different methods of addressing this problem have been developed over the years. The literature shows that extensive work has been carried out over the years to develop methods of optimising mechanism kinematics, generating mechanism designs which can accurately follow given output paths.

When a mechanism is actuated at high speeds, a range of problems are encountered as physical imbalances in the mechanism design are excited, leading to the induction of high frequency harmonic content in its output motion. This harmonic content can greatly limit the maximum operating speed of a mechanism, resulting in loss of motional accuracy and even physical failure of the mechanism. The vast majority of mechanism synthesis methods consider only mechanism kinematics with mechanism dynamics rarely given much consideration, if any at all. The literature indicated that a link exists between the amount of harmonic content present in the output motion of a mechanism and the peak-to-peak magnitude of the variation in drive torque needed to generate that motion.

A prototype servo-mechanical test rig was provided by the project's industrial sponsor company, ITCM Ltd. The test rig possessed a multi-link mechanism, nicknamed the Woodpecker mechanism, which had been observed to have an output motion rich in harmonic content. This mechanism was modelled and used as a test case for development work. A reconfigurable test rig was also designed and built. Using this test rig, it was possible to assemble and actuate a wide variety of mechanism designs without re-fabricating parts.

The work described in this thesis details the development of two contrasting approaches of optimising the performance of servomechanisms for operation at high speeds. The output motions of these mechanisms will possess low levels of harmonic content.

The first method was entitled the Cam Function Generation Method. This method is applicable for use with existing servomechanisms. Using this method, a variable velocity demand signal was generated. This demand signal takes into account and compensates for the dynamic characteristics of the system. The resultant motion was simulated to yield reductions in peak-to-peak magnitude of drive torque magnitude. The method was applied to the prototype test rig. The experimental results confirmed the simulated results.

The second approach tackles the problem of synthesising new mechanism designs taking into account mechanism dynamics and kinematic to generate mechanisms which possess not only good motional accuracy but also good high speed operating characteristics. A mechanism synthesis tool called SWORDS was used to perform the kinematic synthesis whilst a dynamic simulation program entitled DYSIM was used to perform the dynamic analysis using inverse dynamics. The process was applied to the Woodpecker mechanism. A series of three superior alternative mechanism designs were synthesised and a hierarchy of effectiveness between them was identified.

Dynamic models of the reconfigurable test rig, configured in the form of the seed mechanism and the three alternative mechanism designs were created. Inverse dynamic analysis revealed a change in performance hierarchy compared to that identified during the synthesis process. The seed mechanism and the alternative mechanisms were assembled and actuated using a velocity step demand signal to analyse their dynamic qualities. The experimental results verified the simulation results with the same hierarchy of performance being observed. The Cam Function Generation Method was applied to the seed mechanism and the dynamically most superior alternative mechanism design. It was possible to observe that the alternative mechanisms possessed considerably better dynamic characteristics than the seed mechanism. The Cam Function Generation method was applied to the most superior alternative mechanism design. Additional dynamic improvements were observed.used to characterize the surge pressure and volume overshoot across a range of test conditions, the characterized surge pressure being within 6% of the measured values.


Linear decentralised modelling for H∞ control of a multi-axis simulation table

Ali Ghorashi, 2010


In this thesis a broad review of the existing control methods for hydraulic systems is conducted and advantages and shortcomings of each method are discussed. With the focus on stability and performance robustness, the H∞ control approach is chosen as a reliable method for the control of multi-axis hydraulic test rigs.

An improved decentralised linear modelling method for multi-axis electro- hydraulically actuated servo systems is proposed. The model is experimentally validated for a Multi Axis Simulation Table (MAST), with 6 degrees of freedom, located at the University of Bath. The dynamic analysis of the MAST confirms the superiority of the modelling technique compared with conventional methods, in terms of deriving a close linear fit to the physical non-linear plant.

Based on the validated linear model the uncertainties of the system are quantified, which leads to the design of an H∞ controller for the X axis (direction) of the MAST. The performance of this controller is then compared with that of a classical Proportional-Integral-Derivative-lag (PIDL) controller.

Furthermore, the control design is extended to that for the table coupled with parameter varying specimens. A mathematical model for the specimen is presented, however, the adoption of a pragmatic black box modelling approach for the assessment of dynamic interactions between the table and the specimens is then justified. This leads to the H∞ controller design for the coupled table and specimens. The performance of the H∞ controller is assessed for different types of specimen coupled with the MAST. Moreover, the H∞ controller performance is compared with that of the conventional PIDL control method for the different scenarios.

To show the effectiveness of the H∞ controller for Laboratory Dynamic Testing (LDT), a demonstration of an LDT is presented. A comparison between the performance of the H∞ controller and that of the PIDL controller for this case is also presented.


The application of self-learning to injection moulding machines

Dimitris Fronimidis, 2007


Production of quality injection moulded parts is a complex task that requires a deep understanding of the interaction between machine settings and in-mould parameters. This project reports on the monitoring of the polymer dynamics during the injection cycle and proposes an effective control scheme for the process.

The study is focused on an Arburg 25-tonne injection moulding machine which is hydraulically-actuated. For the modelling and simulation of the filling and packing phases, the dynamics of both the machine's hydraulic circuit and the polymer (polypropylene) behaviour were investigated. The simulations were validated on a modified version of the injection moulding machine in which a specially instrumented mould was used.

To assess the extent of solidification of the part and identify phase changes during the cycle, two monitoring methods were studied. One makes use of ultrasound transducers while the other utilizes fast-response thermocouples. Both methods were found to enhance the control of the process. The ultrasound feedback provided sufficient information for quick set up of the controller in real time.

A hybrid minimal controller synthesis (MCS) controller was developed and evaluated experimentally for the closed-loop control of flow and pressure trajectories. The algorithm does not require a priori information about the plant dynamics. To reduce the MCS sensitivity to noise in the feedback signals, a modification of the MCS is proposed and validated. This approach is shown to enhance the performance of the machine.

A major disadvantage in conventional moulding is the difficulty in influencing the molecular orientation at the core. Vibration of the melt polymer has been applied by previous researchers, by means of additional injection cylinders, because control of fast-acting screw dynamics could not be achieved with conventional control methods. A new method is proposed here, where vibration of the screw in a conventional moulding machine is controlled by the hybrid MCS algorithm.

The mechanical properties of tensile specimens produced with vibration were compared with parts produced by conventional moulding. They show significant improvements; part warpage is reduced by up to 30% and tensile modulus is increased by around 10%.


A wavelet based approach to the transient control of rotor/active magnetic bearing systems

Iain Cade, 2006


Magnetic bearings exist in a wide variety of industrial applications of rotating machines. They offer advantages over normal passive bearings with their ability to vary force applied to the rotor and are commonly used with a control strategy to attenuate vibrations levels during operation.

This thesis is focused on the use of wavelet analysis in detecting and controlling the response due to disturbances in flexible rotor/magnetic bearing systems. A method for detecting sudden changes in synchronous forcing and rotor/auxiliary bearing contact is presented using wavelet coefficients. Artefacts associated with mass-loss and rotor/auxiliary bearing contact are identified in simulation and through experiment and shown to be distinguishable.

The discrete time transfer function of a rotor dynamic system is transformed to consider the behaviour of signal wavelet coefficients. This provides the basis for system analysis and controller design in the wavelet domain. A method of steady state identification is demonstrated using wavelet analysis to determine the long term behaviour of individual wavelet coefficient levels. Experimental validation is shown to identify the steady state vibration with a reduced transient response. Controller stability and performance are demonstrated using steady state prediction to moderate control signals.

A novel method for transient rotor/active magnetic bearing control using sampled wavelet coefficients is proposed. The wavelet based controller is designed from target transient responses due to step changes in wavelet coefficients of applied forces. Transient system dynamics are embedded in the controller and evaluated from on-line system identification. Simulated and actual mass-loss tests were performed at critical speeds corresponding to near sudden changes in unbalance that are capable of exciting rotor dynamic modes in a transient manner. The controller is shown to suppress the transient responses within a finite settling time.


Dynamic simulation and high-speed precision control of spatial manipulators

Yanzhi Li, 2007


In the past 20 years, parallel manipulators have drawn many interests in robotics research for their potential features such as high accuracy, rigidity, speed and large load carrying capability. An increasing number of papers have been published on this subject with applications in very different domains. However, it is well known that, in practice, some interesting features have potential, but many unexpected difficulties in design and control have led to performance below expectation.

The purpose of this thesis is to investigate dynamics and control of high performance parallel manipulators. It focuses on the control of parallel manipulators with motion planing, kinematics, dynamics, control arithmetic and software implementation in five crucial parts. Two models are selected for simulation purposes, one is a 3 DOF (Degree of Freedom) RRR (Revolute Revolute Revolute) PPM (Planar Parallel Manipulator) and the other is the 6 DOF MAST (Multi Axis Shaking System).

Being one of the most popular methods to solve dynamic problems for complicated mechanisms, Lagrangian dynamics are chosen as the tool to build dynamic models for the parallel manipulators. Dysim is selected as the software package to construct the numerical models of parallel manipulators. Models are used on the Matlab and Simulink environments through Dysim’s Simulink toolbox sDysim. The motion trajectory is defined by exponential functions. Also, it assumes the desired trajectory is within the physical workspace of the manipulator and is far away from singular points. A Schroeder Phased Harmonic Sequence (SPHS) is formulated as the external disturbance for the system identification and control purposes.

According to the inverse dynamics analysis for the selected motion trajectories, a set of independent second order systems for parallel manipulators is achieved in terms of linearized system models by using a system identification technique, based on least squares estimation. Stability and reliability of the linearization of the system model are examined under different conditions by statistic measures, such as Goodness of Fit (GOF) and Standard Error (SE). A piecewise linearization method is also introduced to improve the estimation result in the case of lower GOF. The same technique is used to obtain a linearized error model.

An error model is presented in order to investigate significant error dynamics characteristics for parallel manipulators. Other than the uncoupled system error model mentioned, the error model is constructed in consideration of the coupling between different limbs. The detailed description of this method along with mathematical examples is explained in Chapter 4.

A scaled SPHS disturbance is selected as the unknown noise added to tested models to examine control algorithms. A Linear quadratic regulator (LQR) is chosen as the controller based on the error models obtained while Bryson’s rule gives direction to find appropriate weighting matrices. The Kalman state estimator is unnecessary since the error model provides full state measurements by the inverse dynamics analysis. Depending on system characteristics and control effect, weighting matrices are changed manually. In the case of worse estimation of error models, an integrator is added in parallel to the LQR. Different control algorithms are applied on a RRR PPM and the MAST models and results are presented and discussed.


Robust control and contact recovery of rotor/magnetic bearing systems

Michael Schlotter, 2007


A basis for modern controller design is an accurate model of the plant. A detailed finite element modelling (FEM) framework for rotor/magnetic bearing systems is presented, and the derived FEM model of the flexible rotor test rig is then used for robust H∞ control design. A novel, versatile LMI based design method is introduced, which takes the physical structure of the system into account. This facilitates the specification of design criteria and allows for the formulation of multiple design objectives in one unified framework. Conventional linear time invariant (LTI) H∞ controllers are designed for one particular rotational speed only. When applied to different speeds, plant variations may lead to closed loop instability. The applied LMI gain-scheduling techniques, however, guarantee robust stability when the system operates at varying rotational speeds. All derived H∞ controllers are tested by simulations and experiments. Their stability and performance is examined and compared with decentralised proportional-derivative (PD) control.

In the second part of this thesis, rotor-stator contact dynamics are investigated, and new control strategies for contact recovery are developed. It is shown that rotors subject to constant synchronous excitation can adopt stable periodic contact modes. For recovery, synchronous forces are calculated based on the contact dynamics model, and applied through the magnetic bearings. These forces reduce the excitation causing the contact to a level at which the modes become unstable and the rotor can progress to a contact-free orbit. The procedure is demonstrated by simulations of a simple disk system before it is applied to an experimental flexible rotor test facility. Error margins are investigated and possible limitations are discussed. Although the method is developed for PD controlled systems only, it is shown experimentally that it can also be successfully applied to H∞ controlled rotors.


Modelling and control of contact in magnetic bearing/flexible rotor systems

Abdul-Hadi Abulrub, 2006


Active Magnetic Bearings (AMBs) have industrial interest in high-speed machinery due to their advantages of no mechanical contact, no need for lubrication, high precision operation, and ability to operate at high rotational speeds. An AMB levitates a rotor by means of electro-magnetic forces at an equilibrium position at which all forces acting on the rotor are balanced. The principle of active control is to measure the rotor position; any deviation will influence the magnetic flux, which is then altered to control the rotor position. Despite many inherent advantages, active magnetic bearings suffer from limited force capacity. This restricts their applications when sudden external impacts are expected, or when a sudden change of unbalance occurs. in these cases, the rotor may make contact with an auxiliary bearing, which introduces non-linear dynamics. This thesis details research undertaken to investigate the behaviour of a rotor in an AMB system when contact occurs with auxiliary bearings. Several techniques have been reported to examine the contact dynamics modelling, most of them represented the contact dynamics with non-linear stiffness and damping elements. These characteristics lead to computationally inefficiency. A possible approach to improve matters is to use Lagrange’s energy method and introduce constraints on the system. The constrained Lagrangian equations of motion do not require the modelling of the contact forces. When contact occurs, a constraint on the generalised coordinates and the constrained equations are added to the equations.


Modelling and control of multi-body structures actuated through large displacements

David Branson III, 2006


Current trends for modern structures are towards complex and lightweight designs that may involve some form of actuation. Although this has produced structures that are visually stunning, the increased flexibility has made them more susceptible to disturbances caused by external forces, such as wind loads, or the actuation system used to move the structure. Currently, vibration suppression in large structures tends to be passive in nature. However, there are many advantages to be gained from using active control in the suppression of structural vibration. To assist in the design of suitable active damping and position control methods an accurate model of the structure is required that can be readily coupled to a non-linear dynamic model of an actuation system.

This thesis describes the research undertaken on the control of multi-body structures actuated by hydraulic systems. An experimental test rig was designed and built to include a structure with hydraulic actuation of rigid and flexible structural components. Models of the hydraulic components were based on experimental and theoretical data, and these were combined with a finite element based structural model that coupled the response of free-free flexible modes to rigid body motions.

The full simulation model was validated by comparison to open-loop experimental results, and then used in the development of controllers based on Proportional-Integral (PI) feedback and H¥ optimisation. Experimental evaluation of the controllers found that for a variety of operating conditions:

  • H∞ control methods result in stable and more accurate position control of the rigid components when compared to PI feedback
  • The developed Single-Input/Multiple-Output (SIMO) H∞ controller was effective at attenuating vibrations in the flexible components of the structure without reducing the desired position control accuracy.

The methodology and strategies designed can be used to improve position control, and reduce unwanted vibrations in multi-body structures such as bridges, buildings, robot linkages, and space structures.