Introduction
Lecture 1
The lecture had three parts.
Part I Course Philosophy
Tried to indicate the importance of the integration of knowledge from previous courses to contruct mathematical models of chemical processes. The use of kinetics, transport phenomena, thermodynamics etc is paramount to the successful modeling of processes from first principles.
The key difference between the courses in the senior year and the junior and sophomore years is the shift from analysis to synthesis. This is not to say that NO synthesis is done in the earlier courses, but that has not been the focus.
Part II Course Outline
The overall flow of the course is as given below:
The basic phases of the course are modeling, which leads to a set of non-linear coupled ode's, linearization and deviation, which leads to a set of coupled linear ode's, Laplace transforms, which leads to transfer functions, and then controller design and analysis. The course ends with looking at multivariable control and model predictive control, and some ideas on the interaction of design and control.
Part III Historical Perspective
As described there appear to be two underlying themes to the development of process control and three important periods.
The 1950's saw the naive adoption of certain set point tracking approaches to control developed to point guns (important - wrongly pointed guns are a serious hazard - almost as serious a hazard as correctly pointed ones - the issue is the subject of the hazard) to process control. This view was recognized as leading to poor control because maintaining all the process variables at their set points is often not feasible and leads to problems with the propagation of disturbances. The view of the system did evolve from "instruments + hardware" into "instruments + hardware + process."
The 1960's saw the advent of computer control and the use of digital control analysis. In these early years the issue of sample and control intervals (i.e. only being able to update on an infrequent basis) caused problems, these have largely been eliminated because of the improvements in speed.
The 1970-80's saw the development of the model predictive control strategy, first at Shell for the control of large scale complex processes. The key advantages of MPC were its inherently multivariable nature, its ability to handle constraints, and its use of an objective function to put the response to disturbances and setpoint changes on a consistent basis.
In the 1990's MPC became widely adopted as vendors developed "turn key" solutions for this technology.
The themes of the development are really twofold
- Sensors Process control is helped tremendously by the ability to perform better sensing of process variables such as composition. Better means, faster, more accurate and precise, cheaper, more reliable, and fundamentally different. For instance the ability to sense the DNA of someone enables the design of control schemes for diet, exercise etc that can help control diseases such as diabetes.
- Computers Process control is increasingly reliant on process models. The ability to compute the solution to large sets of equations rapidly and reliably is aided by the advances in computer hardware. Since the late 1940's the growth in computer power has been exponential and it is our ability to construct models that is probably the most significant limitation for process control applications.
For an example of the growth of computing power consider the deep blue program at IBM, their latest challenge - the protein folding problem.
http://www.research.ibm.com/news/detail/bluegene.html
and their general "deep computing" initiative:
http://www.research.ibm.com/dci/cat1_compute.html
The Future
In the future we can expect these trends to continue to play a role in process control. In particular advances in sensors will be fueled by micro-fabrication technologies, leading to cheaper, faster sensors. Our understanding of biological phenomenon and their use to produce direct sensing of biological state will also be important factors. Advances in computing will not have significant impact on process control unless these are mediated by better phenomenological understanding of processes or by our ability to construct empirical models that are more sophisticated and yet come with reasonable guarantees of performance.
Introduction Self-Study
Links to this Page