Data processing: generic control systems or specific application – Generic control system – apparatus or process – Plural variables
Reexamination Certificate
1998-10-06
2002-02-19
Grant, William (Department: 2121)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Plural variables
C700S028000
Reexamination Certificate
active
06349236
ABSTRACT:
This invention relates to the field of process control systems and, in particular, to a method for adaptively relinquishing and reasserting control over selected controlled variables in a multivariate control system.
BACKGROUND
In a process control system, a controller compares values of a process characteristic, or “controlled variable,” with a target value, or “setpoint,” to determine whether the process is operating within acceptable bounds. For example, in a biscuit baking process, a controller compares the moisture of the baked product with a desired moisture content value to determine whether batches of biscuits are properly baked.
If the controller determines that the controlled variable is outside normal bounds, it can initiate corrective action by adjusting a parameter, or “manipulated variable,” of the process. In the biscuit baking example, if the sample is too moist, the controller can signal an increase in temperature.
In the control system described above, there is but a single controlled variable, moisture, and a single manipulated variable, temperature. Mathematically, the above control system can be represented by the equation
u=f(r, y)
where y represents the controlled variable, moisture, u represents the manipulated variable, temperature, r represents the setpoint, and f represents the system transfer function.
In a multivariate control system, there can be several controlled variables and several setpoints, one corresponding to each controlled variable. Additionally, there can be several manipulated variables. Thus, a multivariate system to be controlled can be characterized by a system transfer function {overscore (f)} such that
{overscore (u)}={overscore (f)}({overscore (r)}, {overscore (y)}) (1)
where {overscore (u)} is a vector of manipulated variables, {overscore (r)} is a vector of setpoints, and {overscore (y)} is a vector of controlled variables. To control such a system, it is necessary to choose {overscore (u)} to minimize the error between the vector of controlled variables, {overscore (y)}, and the vector of setpoints {overscore (r)}. These setpoints represent the desired values for the controlled variables.
In the biscuit baking system described above, if the moisture sensor malfunctions, or if the temperature is already as high as it can go, the control system can no longer function effectively. The situation is different in a multivariate system. If, for example, the device measuring one of the controlled variables malfunctions, there remain other controlled variables which can be controlled. Conversely, if for some reason one of the manipulated variables can no longer be used, there remain other manipulated variables which can be used to control the controlled variables. In either case, it may still be possible to maintain control over some, but not all, of the controlled variables.
Equation (1) provides a mathematical basis for the loss of control over at least one controlled variable upon the loss of a manipulated variable. As a general rule, a unique solution to equation (1) exists only when the dimensionality of the manipulated variable vector {overscore (u)} equals the dimensionality of the controlled variable vector {overscore (y)}. Where there are more controlled variables than there are manipulated variables, it is generally not possible to control each controlled variable.
In order to maintain control of the multivariate control system, it can be necessary to relinquish control of at least some of the controlled variables. This process of rendering a controlled variable inactive generally includes a step of assigning a priority order to each controlled variable. In this way, control over those controlled variables which have lower priority can be relinquished before control over those controlled variables which have higher priority.
It is known in the art to assign a priority value to each controlled variable and to relinquish control over the controlled variable having the lowest priority value. For example, in a refrigerated boxcar in which one can control both temperature and humidity, one may be willing to forego control of humidity if the alternative is to lose control of the temperature. In such a control system, the humidity has a lower priority value than the temperature and hence control over humidity is relinquished before control over the temperature.
A typical method for selecting those controlled variables over which control is to be relinquished includes the steps of counting the number of active controlled variables and the number of available manipulated variables. By comparing these two numbers, a prior art controller can determine how many controlled variables must be rendered inactive. Typically, the controlled variables are arranged in a pre-determined priority order which stays fixed throughout the interval during which the system is to be controlled. The processor then relinquishes control over controlled variables in the order specified by this predetermined priority order, beginning with the lowest priority controlled variable.
A typical implementation of the above prior art method includes the step of counting the number of active controlled variables in a vector of controlled variables by determining whether the measurement of each controlled variable is a valid measurement. If the measurement is valid, the controller considers that controlled variable to be an active controlled variable. If the measurement is invalid, an attempt to exert control over that controlled variable is pointless. Accordingly, the controller relinquishes control over that controlled variable by setting its value equal to its last valid measured value and recalculating the gain matrix of the controller so that the gains associated with that value are zero. This method of relinquishing control over a controlled variable thus suffers from the disadvantage that the gain matrix of the controller is recalculated when the measured value of a controlled variable becomes invalid.
The prior art method also includes the step of counting the number of manipulated variables available for controlling the controlled variables. This includes the steps of determining whether each manipulated variable is output limited. An example of an output limited manipulated variable is one associated with a valve that is already fully open or fully shut. If a particular manipulated variable is not output limited, the controller increments a counter by one. Then, the controller proceeds to the next manipulated variable, if any.
Having determined the number of active controlled variables and the number of available manipulated variables, the prior art method then determines whether the difference between the number of active controlled variables and the number of available manipulated variables requires adjustment. If the number of active controlled variables exceeds the number of available manipulated variables, the controller relinquishes control over the lowest priority active controlled variable by recalculating the gain matrix of the controller so as to render the gains associated with that controlled variable equal to zero. This process is repeated until the number of active controlled variables and the number of available manipulated variables is such that equation (1) has a solution.
It is apparent that the prior art step of counting active controlled variables does not provide for recovery of a controlled variable once control over it has been relinquished. Consequently, a spurious measurement of a controlled variable may result in permanent loss of control over that controlled variable, even though the measurement of that controlled variable is generally reliable.
It is also apparent that the foregoing method relinquishes control over controlled variables according to a predefined priority with no consideration given to the measured values of the remaining controlled variables. The foregoing method is, as a result, not adaptive. A disadvantage of such a non-adaptive method is that it is possible to relinquish control over a lo
Di Vittorio Nicholas P.
Hansen Peter D.
Foley Hoag & Eliot LLP
Frank Elliot
Grant William
Liepmann W. Hugo
Oliver Kevin A.
LandOfFree
Adaptive shed strategy for constraint enforcement in a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptive shed strategy for constraint enforcement in a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptive shed strategy for constraint enforcement in a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2979745