E-News December 2015 Edition

New Trends in the Delivery of In-House Training Courses by Professor Mike J Grimble

The most popular training courses developed by the ACTC program and provided by ISC are those concerned with Control Fundamentals or an Introduction to Control. They are aimed at engineers that only have a superficial knowledge of control engineering, or where they perhaps attended courses some years ago and have been working in other areas. They are often used to build up the knowledge of engineers that work with the control specialists to make that work more effective.

At the other extreme more courses are now being produced in very specialist areas such as gasoline or diesel engine controls, wind turbine controls, servo systems, process controls, offshore oil industry controls and signal processing for the aerospace and defence industries. The success of the Rolling Mill Academies some years ago stimulated the development of metal processing courses but the growth in other areas has mainly been in response to particular demands.

The standard courses on subjects ranging from robust control to optimisation have also learned from this experience. More of the standard courses now include detailed design examples, where one problem is used to illustrate many features in control design.

The fast majority of training courses are provided at company premises, often held in the USA or mainland Europe. For example, over the last few months courses have been presented to Cummins Diesel Engines in Columbus, Indiana, to Ford in Dearborn, Michigan and a state of the art workshop has been organised with General Motors in Warren, Michigan.

If you have an interest in training courses please contact Dr Meghan McGookin

Mike Grimble

Professor Mike Grimble and Dr Andy Clegg Speak at the IET on "Use of dynamic simulation in the development of control systems".

On 3rd September an audience of more than 30 people got together in Teacher Building, Glasgow to hear Dr Andy Clegg and Professor Mike Grimble deliver presentations on the "Use of dynamic simulation in the development of control systems".

The audience learnt that design, development and deployment of control systems can often be challenging; especially in time, cost or safety critical applications, such as in oil & gas, maritime or nuclear industries and how the use of dynamic simulation of plant/control systems can help.

Professor Mike Grimble from the University of Strathclyde presented a state-of-the-art review of the use of modelling and simulation in the design of industrial applications from diesel engines to marine system controls.

Watch Mike Grimble's presentation here

Dr. Andy Clegg from Industrial Systems and Control Ltd talked about his experiences with developing and using simulations within the design and deployment for challenging machinery control and process control applications. He covered how simulation was used to shrink commissioning time for marine hydraulic control systems and ensure things are right first time.

Watch Andy Clegg's presentation here

It was an excellent event, with a very engaged audience, if the number of questions at the end was any measure!


H-infinity Robust Control – Recent Developments

Since H-infinity robust control was introduced by researchers such as Zames, Francis, Glover, Doyle, Athens and Safonov, amongst others. It had a relatively slow growth after the initial excitement and intense activity. It seemed that once the algorithms were available to easily calculate such control laws the academic communities interest diminished. However, in more recent years researchers such as Isidori and Helton have developed nonlinear versions which reinvigorated academic activity. During this time there have been many more application papers demonstrating robustness improvements can be achieved.

One of the early papers which initiated interest in the subject was the contribution by Doyle which showed LQ regulators have no guaranteed robustness properties if states are not available for feedback. The introduction of the Kalman filter in an LQG solution means that the guaranteed properties are lost. The example he used to demonstrate the problems with LQG controls was however very special involving an open-loop unstable system with a repeated root at s = 1. Even the cost-function had an unusual and very suspect state-weighting matrix. This did not of course detract from his main point which was to show there are no guaranteed margins with LQG controllers, but it did leave the question open of whether H-infinity would be that much better than LQG when weightings were particularly selected to try to achieve robustness. Such an assessment does of course depend of the type of uncertainty to be expected and its structure.

Whether H-infinity design is that much more robust than LQG does therefore depend upon the application but it certainly has some advantages. For example, it directly minimises sensitivity functions, which are intimately related to robustness properties. It also allows uncertainty described to be pre-specified and for the optimal robust solution to be achieved (for certain classes of problem). Moreover, it has some unexpected properties. That is, it provided excellent step-responses and non-interacting multivariable solutions.

The current trend is to use H-infinity design for nonlinear systems by employing the well-established technique of scheduling. The idea is to account for the uncertainty of moving between different nonlinear regions of operation by exploiting the H8 robustness capabilities. There have been good reported results in application studies and this seems a reasonably practical approach. Full nonlinear H-infinity design is still more of a mathematical research topic, but it may have some benefits for future applications. This is therefore an area to watch and perhaps even evaluate when robustness is the key issue and problem.


Non-linear Predictive Generalised Minimum Variance State-dependent Control

Research Article by Professor Mike J Grimble and Dr Pawel Majecki

Abstract: A non-linear predictive generalised minimum variance control algorithm is introduced for the control of nonlinear discrete-time state-dependent multivariable systems. The process model includes two different types of subsystems to provide a variety of means of modelling the system and inferential control of certain outputs is available. A statedependent output model is driven from an unstructured non-linear input subsystem which can include explicit transportdelays. A multi-step predictive control cost function is to be minimised involving weighted error, and either absolute or incremental control signal costing terms. Different patterns of a reduced number of future controls can be used to limit the computational demands.

Download the full article here


IEEE Industrial Applications Society

The IEEE Industrial Applications Society is 50 years old and the UKRI IAS Chapter is Co-Chaired by Prof Amir Hussain (Stirling University) and Prof Mike Grimble (Strathclyde University). A very successful recent meeting was to set up a Joint Industry Applications Society / Power Electronics Society Student Branch Chapter. The UK and Republic of Ireland (UKRI) Industry Applications Society (IAS) Chapter probably has one of the broadest scopes, it is primarily aimed at exploiting industrial applications of Electrical and Electronic Engineering (EEE) technologies.

The chapter is currently, focussed on exploring future prospects in some of the most challenging IAS technical areas of interest - primarily the development and industrial exploitation of multi-disciplinary, control, signal image processing and brain-inspired computational intelligence technologies, to engineer the intelligent and engineering systems of tomorrow! The chapter actively fosters mutually productive links between academia and industry, both nationally and internationally. If you wish to contribute contact Amir Hussain or Professor Mike Grimble.


Optimizing the Tuning of Controllers

Software is available to enable the best tuning of a controller to be determined and for recommended tuning parameters to be generated. The optimization algorithms are sometimes called restricted structure optimal control laws. The restricted structure controllers might for example be of PID or extended PID structures. This software also enables the controller performance to be benchmarked.

The decision as to what is good controller tuning does of course depend on the requirements of the process. The software tools normally use measures of performance that are based on an optimal control cost-function. Thus, for example, if the variance of the controlled quantities are of interest a minimum variance criterion would be used as the benchmark. However, there are vast ranges of other performance metrics which can be introduced, depending upon company needs. A predictive control benchmarking method has for example been developed.

These techniques are sometimes used just to assess how well controllers are behaving and this is of course the subject of performance benchmarking. It is a step further for the best set of controller parameters to be provided, and it depends upon the benchmarking tool available whether this is possible or not. Most major SCADA manufacturers utilise this type of software but it can also be provided in a more bespoke form for particular applications. If you are interested in the subject of restricted structure optimal control and performance benchmarking you may contact Professor Mike J Grimble


IFAC Conference on Nonlinear Model Predictive Control 2015, Seville

The 5th IFAC Conference on Nonlinear Model Predictive Control 2015 (NMPC'15) was held in Seville, Spain, September 17 - 20, 2015. The plenary talk by Rolf Findeisen was control in the era of the internet and communications. He explained about the variety of work there now is in areas such as networked control systems, cyber physical systems, wireless sensor networks and others. His main message was that predictive control was well suited to handle many of the challenges now appearing in this area.

There was an excellent industrial session with contributors such as Richard Braatz of MIT. He described the application of MPC to total process control. He explained such systems can have high dimensions (greater than 10,000 states), and they are modelled by a combination of differential and algebraic equations. There is one simplification and that is such systems are often stabilised by the design of the lower level PI or PID control loops. The upper level optimisation is therefore on a stable process. However, many states are not observable. Such systems are modelled by differential/algebraic equations and there is a high cost of first principles modelling.

Braatz (MIT) noted that unfortunately many chemical plants are individual constructions and unique. This is unlike areas such as automotive systems, where MPC is highly leveraged by the application to a large number of manufactured products. He suggested that input-output approaches are often more pragmatic for modelling such systems. He also observed that there is a need for hybrid system controls, since such plants go through start up, shut down and warm start arrangements. There is also a need for fault tolerant control for large nonlinear MPC systems.

In an industrial session, Stephano Di Cairano (Mitsubishi Electric Research Laboratories), described the large range of applications. These included air conditioning, automotive equipment, robotics, information and communications, transportation systems and space systems. Research activities involved constraint satisfaction, optimisation algorithms, stability and nonlinear MPC performance. He mentioned they need real guarantees on stability and performance to be provided and research support from the academic community is still required for this area.

One of the presentations in the industrial session was by Lorenzo Fagiano who is an ABB senior scientist. He noted there are 8,500 engineers and technologist and 1,000 of these are in research at their automation centre. He claimed that some of the difficulties of using MPC in process control are the high level of competence in control required and the need to follow up by proper monitoring to ensure performance is maintained over time. He did of course note the many advantages such as constraint handling, MIMO capabilities and the handling of nonlinearities (when using nonlinear MPC).

Lorenzo Fagiano of ABB also discussed so called scenario MPC, in a paper, where the aim is to control the rate of constraint violations. This enables a trade of between constraint violation problems and cost optimisation. There are of course some applications where violating constraints is not too disastrous but it may lead to more wear and tear or damage. However there is of course other areas where the constraint is a hard one that cannot be exceeded such as an actuator mechanical limit.

A paper by Normey-Rico was on the robustness of nonlinear model predictive control but drawing analogies with smith predictor type behaviour. He explained how a filter could be used for robustness improvement by drawing an analogy between the two approaches. Another rather practical paper which is quite important for applications was by Pannocchia who described offset free NPC methods including all the subtleties and applications. He reminded us that the traditional method of augmenting the plant model by integrators at the inputs is not so straight forward if the system contains more outputs than inputs.

We enjoyed the presentation by Edwardo Camaro which illustrated the high potential of solar power particularly in countries like Spain. The control problems involve collector movement which is open loop control and temperature and pressure control systems. David Angeli of Imperial College talked about theoretical advances on economic model predictive control with time varying costs. He described how the economic landscape in applications can vary as fast as the system dynamics.

Mike Huang described an interesting application of nonlinear MPC to diesel engine control whilst working for Toyota. His main interest was in reducing the complexity of algorithms to simplify implementation and reduce computational effort. The main object of his work was to regulate the intake manifold pressure and compressor flow to achieve specified set points by coordinated control of a variable geometry turbine and exhaust gas recirculating valve. It is the limit in resources of the ECU which increases the computational complexity of implementing NMPC. The authors of this paper included a comparative assessment for computational times and constraint violations for different constraint handling techniques.

A poster presentation by Mark Cannon and Raynar Schaich (Oxford), involving linear robust model predictive control using state and input dependant disturbances was particularly interesting. They employed multiparametric quadratic programming (mpQP).

This was an excellent meeting with leading world researchers in the subject and very well organized. The venue (Nh Collection) and City were splendid and the organizers are to be congratulated on a very successful meeting.


Book Review: Book Review: Perspectives on Defence Systems Analysis, by: William P. Delaney, published by: The MIT Press, £65.95

This is a valuable text for those involved with the aerospace and defence industries but it does not cover the design methods or practical systems use, which are more usual. It is written at a higher level than the rather mathematical subjects of control or systems engineering. It provides a broad view of defence systems and their needs. It has many contributions by distinguished authors in the subject and is well illustrated. For those that need to understand the context of their research and development in aerospace and defence systems this is a valuable contribution with a relatively modest cost of £65.95.

Book Nerd.


Book Review: Set-Theoretic Methods in Control, by Franco Blanchini and Stefano Miani, published by Birkhäuser, £70

The second edition of Set-Theoretic Methods in Control by Franco Blanchini and Stefano Miani is published by Birkhäuser and as the title suggests this is a research oriented text. However, it deals with many of the topics that are needed in more engineering design areas, covering topics of importance to industry. The cost of the text is £70.

Lyapunov methods are described and when uncertainties are present so called polytopic system models are introduced. Many of the topics that are used in predictive control and robust control applications are covered such as convex sets, polyhedral sets and related topics such as polytopes. There is a whole chapter on dynamic programming which is important in all areas of optimal control, and rather practical subjects like robust stability radius are covered.

Linear parameter varying systems now feature in many papers on both linear and nonlinear systems and the results are full chapter on this topic. Such models are used on everything ranging from automotive powertrain controls to wind turbine controls. There is a rather unusual chapter on control of systems with time domain constraints. Constrained handling is of course important whether it be input constraints or possible state constraints.

One of the modern trends in advanced control is towards a greater use of hybrid systems theory and switching, and both of these are covered in a chapter. Throughout the text there are examples and problems. For example, in the section on switching a junction traffic control problem is considered. A chapter is also included on suboptimal control methods and this includes constrained receding horizon control which is of course very popular.

Worst case estimation methods are also covered and different application areas such as communication and network problems. This is a surprising text in some ways. From the title I expected a more narrowly focused text on rather abstract mathematic al methods but in fact it is better to consider this as a summary of mathematical tools that are needed in a lot of engineering control problems. The material is very nicely presented and written.

Book Nerd


Book Review: Sliding Mode Control, The Delta-Sigma Modulation Approach, by Hebertt Sira-Ramírez, published by Birkhäuser, £76.50

The new book by Hebertt Sira-Ramírez published by Birhauser is on Sliding Mode Control: The Delta-Sigma Modulation Approach. The hardcover version of the book is £76.50 but this can also be purchased as an e-Book which costs £60.99. Sliding mode control is one of these topics which is migrated from very theoretical background studied by researchers in the former Soviet Union to become a mainstream research topic with good applications potential.

This particular text considers switch-regulated nonlinear systems. There are numerous references to applications throughout the text which has a clear and logical presentation of ideas. There is a lot of interest in switching system mainly driven by interest from application areas and hence this text is rather topical. Applications such as automotive motion control, double bridge buck convertors, space vehicles and water tank controls are all considered.

This book is probably more aimed at advanced graduate level courses in control engineering or researchers in University departments. However, it is a valuable addition to anyone wishing to use sliding mode control methods.

Book Nerd


Industrial Monograph Series

Professor Mike Grimble can be contacted if you have ideas for an Industrial Monograph. The series with editors Grimble and Johnson is entitled Advances in Industrial Control. It is a series of monographs focussing on the applications of advanced and novel control methods. The series has worldwide distribution to engineers, researchers and libraries, and promotes the exchange of information between academia and industry. The books all demonstrate some theoretical aspect of an advanced control methods and show how it can be applied either in a pilot plant or in some real industrial situation. Note that "industrial" here has a very broad interpretation; it applies not merely to the processes employed in industrial plants but to systems such as avionics and automotive drivetrains.