Method and apparatus for controlling a non-linear mill

Abstract

A method and apparatus for controlling a non-linear mill. A linear controller is provided having a linear gain k that is operable to receive inputs representing measured variables of the plant and predict on an output of the linear controller predicted control values for manipulatible variables that control the plant. A non-linear model of the plant is provided for storing a representation of the plant over a trained region of the operating input space and having a steady-state gain K associated therewith. The gain k of the linear model is adjusted with the gain K of the non-linear model in accordance with a predetermined relationship as the measured variables change the operating region of the input space at which the linear controller is predicting the values for the manipulatible variables. The predicted manipulatible variables are then output after the step of adjusting the gain k

Description

TECHNICAL FIELD OF THE INVENTION

[0002] This invention pertains in general to modeling techniques and, more particularly, to combining steady-state and dynamic models for the purpose of prediction, control and optimization for non-linear mill control.

BACKGROUND OF THE INVENTION

[0003] Process models that are utilized for prediction, control and optimization can be divided into two general categories, steady-state models and dynamic models. In each case the model is a mathematical construct that characterizes the process, and process measurements are utilized to parameterize or fit the model so that it replicates the behavior of the process. The mathematical model can then be implemented in a simulator for prediction or inverted by an optimization algorithm for control or optimization.

[0004] Steady-state or static models are utilized in modem process control systems that usually store a great deal of data, this data typically containing steady-state information at many different operating conditions. The steady-state information is utilized to train a non-linear model wherein the process input variables are represented by the vector U that is processed through the model to output the dependent variable Y. The non-linear model is a steady-state phenomenological or empirical model developed utilizing several ordered pairs (Ui, Yi) of data from different measured steady states. If a model is represented as:

Y=P (U, Y)  (001)

[0005] where P is some parameterization, then the steady-state modeling procedure can be presented as:

({right arrow over (U)},{right arrow over (Y)})→P  (002)

[0006] where U and Y are vectors containing the Ui, Yi ordered pair elements. Given the model P, then the steady-state process gain can be calculated as: 1$\begin{array}{cc}K=\frac{\Delta P\left(U,Y\right)}{\Delta U}& \left(003\right)\end{array}$

[0007] The steady-state model therefore represents the process measurements that are taken when the system is in a “static” mode. These measurements do not account for the perturbations that exist when changing from one steady-state condition to another steady-state condition. This is referred to as the dynamic part of a model.

[0008] A dynamic model is typically a linear model and is obtained from process measurements which are not steady-state measurements; rather, these are the data obtained when the process is moved from one steady-state condition to another steady-state condition. This procedure is where a process input or manipulated variable u(t) is input to a process with a process output or controlled variable y(t) being output and measured. Again, ordered pairs of measured data (u(I), y(I)) can be utilized to parameterize a phenomenological or empirical model, this time the data coming from non-steady-state operation. The dynamic model is represented as:

y (t)=p(u(t),y(t))  (004)

[0009] where p is some parameterization. Then the dynamic modeling procedure can be represented as:

({right arrow over (u)}, {right arrow over (y)})→p  (005)

[0010] Where u and y are vectors containing the (u(I),y(I)) ordered pair elements. Given the model p, then the steady-state gain of a dynamic model can be calculated as: 2$\begin{array}{cc}k=\frac{\Delta p\left(u,y\right)}{\Delta u}& \left(006\right)\end{array}$

[0011] Unfortunately, almost always the dynamic gain k does not equal the steady-state gain K, since the steady-state gain is modeled on a much larger set of data, whereas the dynamic gain is defined around a set of operating conditions wherein an existing set of operating conditions are mildly perturbed. This results in a shortage of sufficient non-linear information in the dynamic data set in which non-linear information is contained within the static model. Therefore, the gain of the system may not be adequately modeled for an existing set of steady-state operating conditions. Thus, when considering two independent models, one for the steady-state model and one for the dynamic model, there is a mis-match between the gains of the two models when used for prediction, control and optimization. The reason for this mis-match are that the steady-state model is non-linear and the dynamic model is linear, such that the gain of the steady-state model changes depending on the process operating point, with the gain of the linear model being fixed. Also, the data utilized to parameterize the dynamic model do not represent the complete operating range of the process, i.e., the dynamic data is only valid in a narrow region. Further, the dynamic model represents the acceleration properties of the process (like inertia) whereas the steady-state model represents the tradeoffs that determine the process final resting value (similar to the tradeoff between gravity and drag that determines terminal velocity in free fall).

[0012] One technique for combining non-linear static models and linear dynamic models is referred to as the Hammerstein model. The Hammerstein model is basically an input-output representation that is decomposed into two coupled parts. This utilizes a set of intermediate variables that are determined by the static models which are then utilized to construct the dynamic model. These two models are not independent and are relatively complex to create.

SUMMARY OF THE INVENTION

[0013] The present invention disclosed and claimed herein comprises a method for controlling a non-linear plant. A linear controller is provided having a linear gain k that is operable to receive inputs representing measured variables of the plant and predict on an output of the linear controller predicted control values for manipulatible variables that control the plant. A non-linear model of the plant is provided for storing a representation of the plant over a trained region of the operating input space and having a steady-state gain K associated therewith. The gain k of the linear model is adjusted with the gain K of the non-linear model in accordance with a predetermined relationship as the measured variables change the operating region of the input space at which the linear controller is predicting the values for the manipulatible variables. The predicted manipulatible variables are then output after the step of adjusting the gain k.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] For a more complete understanding of the present invention and the advantages thereof, reference is now made to the following description taken in conjunction with the accompanying Drawings in which:

[0015] FIG. 1 illustrates a prior art Hammerstein model;

[0016] FIG. 2 illustrates a block diagram of the modeling technique of the present invention;

[0017] FIGS. 3 a -3d illustrate timing diagrams for the various outputs of the system of FIG. 2;

[0018] FIG. 4 illustrates a detailed block diagram of the dynamic model utilizing the identification method;

[0019] FIG. 5 illustrates a block diagram of the operation of the model of FIG. 4;

[0020] FIG. 6 illustrates an example of the modeling technique of the present invention utilized in a control environment;

[0021] FIG. 7 illustrates a diagrammatic view of a change between two steady-state values;

[0022] FIG. 8 illustrates a diagrammatic view of the approximation algorithm for changes in the steady-state value;

[0023] FIG. 9 illustrates a block diagram of the dynamic model;

[0024] FIG. 10 illustrates a detail of the control network utilizing the error constraining algorithm of the present invention;

[0025] FIGS. 11 a and 11b illustrate plots of the input and output during optimization;

[0026] FIG. 12 illustrates a plot depicting desired and predicted behavior;

[0027] FIG. 13 illustrates various plots for controlling a system to force the predicted behavior to the desired behavior;

[0028] FIG. 14 illustrates a plot of the trajectory weighting algorithm of the present invention;

[0029] FIG. 15 illustrates a plot for the constraining algorithm;

[0030] FIG. 16 illustrates a plot of the error algorithm as a function of time;

[0031] FIG. 17 illustrates a flowchart depicting the statistical method for generating the filter and defining the end point for the constraining algorithm of FIG. 15;

[0032] FIG. 18 illustrates a diagrammatic view of the optimization process;

[0033] FIG. 18 a illustrates a diagrammatic representation of the manner in which the path between steady-state values is mapped through the input and output space;

[0034] FIG. 19 illustrates a flowchart for the optimization procedure;

[0035] FIG. 20 illustrates a diagrammatic view of the input space and the error associated therewith;

[0036] FIG. 21 illustrates a diagrammatic view of the confidence factor in the input space;

[0037] FIG. 22 illustrates a block diagram of the method for utilizing a combination of a non-linear system and a first principal system; and

[0038] FIG. 23 illustrates an alternate embodiment of the embodiment of FIG. 22.

[0039] FIG. 24 illustrates a block diagram of a kiln, cooler and preheater;

[0040] FIG. 25 illustrates a block diagram of a non-linear controlled mill;

[0041] FIG. 26 illustrates a table for various aspects of the mill associated with the fresh feed and the separator speed;

[0042] FIG. 27 illustrates historical modeling data for the mill;

[0043] FIGS. 28 a and 28b illustrate diagrams of sensitivity versus the percent of the Blaine and the percent of the Return, respectively;

[0044] FIG. 29 illustrates non-linear gains for fresh feed responses;

[0045] FIG. 30 illustrates plots of non-linear gains for separator speed responses; and

[0046] FIG. 31 illustrates a closed loop non-linear model predictive control, target change for Blaine.

DETAILED DESCRIPTION OF THE INVENTION

[0047] Referring now to FIG. 1, there is illustrated a diagrammatic view of a Hammerstein model of the prior art. This is comprised of a non-linear static operator model 10 and a linear dynamic model 12, both disposed in a series configuration. The operation of this model is described in H. T. Su, and T. J. McAvoy, “Integration of Multilayer Perceptron Networks and Linear Dynamic Models: A Hammerstein Modeling Approach” to appear in I & EC Fundamentals, paper dated Jul. 7, 1992, which reference is incorporated herein by reference. Hammerstein models in general have been utilized in modeling non-linear systems for some time. The structure of the Hammerstein model illustrated in FIG. 1 utilizes the non-linear static operator model 10 to transform the input U into intermediate variables H. The non-linear operator is usually represented by a finite polynomial expansion. However, this could utilize a neural network or any type of compatible modeling system. The linear dynamic operator model 12 could utilize a discreet dynamic transfer function representing the dynamic relationship between the intermediate variable H and the output Y. For multiple input systems, the non-linear operator could utilize a multilayer neural network, whereas the linear operator could utilize a two layer neural network. A neural network for the static operator is generally well known and described in U.S. Pat. No. 5,353,207, issued Oct. 4, 1994, and assigned to the present assignee, which is incorporated herein by reference. These type of networks are typically referred to as a multilayer feed-forward network which utilizes training in the form of back-propagation. This is typically performed on a large set of training data. Once trained, the network has weights associated therewith, which are stored in a separate database.

[0048] Once the steady-state model is obtained, one can then choose the output vector from the hidden layer in the neural network as the intermediate variable for the Hammerstein model. In order to determine the input for the linear dynamic operator, u(t), it is necessary to scale the output vector h(d) from the non-linear static operator model 10 for the mapping of the intermediate variable h(t) to the output variable of the dynamic model y(t), which is determined by the linear dynamic model.

[0049] During the development of a linear dynamic model to represent the linear dynamic operator, in the Hammerstein model, it is important that the steady-state non-linearity remain the same. To achieve this goal, one must train the dynamic model subject to a constraint so that the non-linearity learned by the steady-state model remains unchanged after the training. This results in a dependency of the two models on each other.

[0050] Referring now to FIG. 2, there is illustrated a block diagram of the modeling method of the present invention, which is referred to as a systematic modeling technique. The general concept of the systematic modeling technique in the present invention results from the observation that, while process gains (steady-state behavior) vary with U's and Y's,(i.e., the gains are non-linear), the process dynamics seemingly vary with time only, (i.e., they can be modeled as locally linear, but time-varied). By utilizing non-linear models for the steady-state behavior and linear models for the dynamic behavior, several practical advantages result. They are as follows:

[0051] 1. Completely rigorous models can be utilized for the steady-state part. This provides a credible basis for economic optimization.

[0052] 2. The linear models for the dynamic part can be updated on-line, i.e., the dynamic parameters that are known to be time-varying can be adapted slowly.

[0053] 3. The gains of the dynamic models and the gains of the steady-state models can be forced to be consistent (k=K).

[0054] With further reference to FIG. 2, there are provided a static or steady-state model 20 and a dynamic model 22. The static model 20, as described above, is a rigorous model that is trained on a large set of steady-state data. The static model 20 will receive a process input U and provide a predicted output Y. These are essentially steady-state values. The steady-state values at a given time are latched in various latches, an input latch 24 and an output latch 26. The latch 24 contains the steady-state value of the input Uss, and the latch 26 contains the steady-state output value Yss. The dynamic model 22 is utilized to predict the behavior of the plant when a change is made from a steady-state value of Yss to a new value Y. The dynamic model 22 receives on the input the dynamic input value u and outputs a predicted dynamic value y. The value u is comprised of the difference between the new value U and the steady-state value in the latch 24, Uss. This is derived from a subtraction circuit 30 which receives on the positive input thereof the output of the latch 24 and on the negative input thereof the new value of U. This therefore represents the delta change from the steady-state. Similarly, on the output the predicted overall dynamic value will be the sum of the output value of the dynamic model, y, and the steady-state output value stored in the latch 26, Yss. These two values are summed with a summing block 34 to provide a predicted output Y. The difference between the value output by the summing junction 34 and the predicted value output by the static model 20 is that the predicted value output by the summing junction 20 accounts for the dynamic operation of the system during a change. For example, to process the input values that are in the input vector U by the static model 20, the rigorous model, can take significantly more time than running a relatively simple dynamic model. The method utilized in the present invention is to force the gain of the dynamic model 22 kd to equal the gain Kss of the static model 20.

[0055] In the static model 20, there is provided a storage block 36 which contains the static coefficients associated with the static model 20 and also the associated gain value Kss. Similarly, the dynamic model 22 has a storage area 38 that is operable to contain the dynamic coefficients and the gain value kd. One of the important aspects of the present invention is a link block 40 that is operable to modify the coefficients in the storage area 38 to force the value of kd to be equal to the value of Kss. Additionally, there is an approximation block 41 that allows approximation of the dynamic gain kd between the modification updates.

[0056] Systematic Model

[0057] The linear dynamic model 22 can generally be represented by the following equations: 3$\begin{array}{cc}\delta y\left(t\right)=\sum _{i=1}^n{b}_i\delta u\left(t-d-i\right)-\sum _{i=1}^n{a}_i\delta y\left(t-i\right)& \left(007\right)\end{array}$

[0058] where:

δy(t)=y(t)−Yss  (008)

δu(t)=u(t)−uss  (009)

[0059] and t is time, ai and bi are real numbers, d is a time delay, u(t) is an input and y(t) an output. The gain is represented by: 4$\begin{array}{cc}\frac{y\left(B\right)}{u\left(B\right)}=k=\frac{\left(\sum _{i=1}^n{b}_i{B}^{i-1}\right)B^d}{1+\sum _{i=1}^n{a}_i{B}^{i-1}}& \left(10\right)\end{array}$

[0060] where B is the backward shift operator B(x(t))=x(t−1), t=time, the ai and bi are real numbers, I is the number of discreet time intervals in the dead-time of the process, and n is the order of the model. This is a general representation of a linear dynamic model, as contained in George E. P. Box and G. M Jenkins, “TIME SERIES ANALYSIS forecasting and control”, Holden-Day, San Francisco, 1976, Section 10.2, Page 345. This reference is incorporated herein by reference.

[0061] The gain of this model can be calculated by setting the value of B equal to a value of “1”. The gain will then be defined by the following equation: 5$\begin{array}{cc}{\left[\frac{y\left(B\right)}{u\left(B\right)}\right]}_{B=1}=k_d=\frac{\sum _{i=1}^n{b}_i}{1+\sum _{i=1}^n{a}_i}& \left(11\right)\end{array}$

[0062] The ai contain the dynamic signature of the process, its unforced, natural response characteristic. They are independent of the process gain. The bi contain part of the dynamic signature of the process; however, they alone contain the result of the forced response. The bi determine the gain k of the dynamic model. See: J. L. Shearer, A. T. Murphy, and H. H. Richardson, “Introduction to System Dynamics”, Addison-Wesley, Reading, Massachusetts, 1967, Chapter 12. This reference is incorporated herein by reference.

[0063] Since the gain Kss of the steady-state model is known, the gain kd of the dynamic model can be forced to match the gain of the steady-state model by scaling the bi parameters. The values of the static and dynamic gains are set equal with the value of bi scaled by the ratio of the two gains: 6$\begin{array}{cc}{\left(b_i\right)}_{\mathrm{scaled}}={\left(b_i\right)}_{\mathrm{old}}\left(\frac{K_{\mathrm{ss}}}{k_d}\right)& \left(12\right)\\ {\left(b_i\right)}_{\mathrm{scaled}}=\frac{{\left(b_i\right)}_{\mathrm{old}}{K}_{\mathrm{ss}}\left(1+\sum _{i=1}^n{a}_i\right)}{\sum _{i=1}^n{b}_i}& \left(13\right)\end{array}$

[0064] This makes the dynamic model consistent with its steady-state counterpart. Therefore, each time the steady-state value changes, this corresponds to a gain Kss of the steady-state model. This value can then be utilized to update the gain kd of the dynamic model and, therefore, compensate for the errors associated with the dynamic model wherein the value of kd is determined based on perturbations in the plant on a given set of operating conditions. Since all operating conditions are not modeled, the step of varying the gain will account for changes in the steady-state starting points.

[0065] Referring now to FIGS. 3a-3d, there are illustrated plots of the system operating in response to a step function wherein the input value U changes from a value of 100 to a value of 110. In FIG. 3a, the value of 100 is referred to as the previous steady-state value Uss. In FIG. 3b, the value of u varies from a value of 0 to a value of 10, this representing the delta between the steady-state value of Uss to the level of 110, represented by reference numeral 42 in FIG. 3a. Therefore, in FIG. 3b the value of u will go from 0 at a level 44, to a value of 10 at a level 46. In FIG. 3c, the output Y is represented as having a steady-state value Yss of 4 at a level 48. When the input value U rises to the level 42 with a value of 110, the output value will rise. This is a predicted value. The predicted value which is the proper output value is represented by a level 50, which level 50 is at a value of 5. Since the steady-state value is at a value of 4, this means that the dynamic system must predict a difference of a value of 1. This is represented by FIG. 3d wherein the dynamic output value y varies from a level 54 having a value of 0 to a level 56 having a value of 1.0. However, without the gain scaling, the dynamic model could, by way of example, predict a value for y of 1.5, represented by dashed level 58, if the steady-state values were outside of the range in which the dynamic model was trained. This would correspond to a value of 5.5 at a level 60 in the plot of FIG. 3c. It can be seen that the dynamic model merely predicts the behavior of the plant from a starting point to a stopping point, not taking into consideration the steady-state values. It assumes that the steady-state values are those that it was trained upon. If the gain kd were not scaled, then the dynamic model would assume that the steady-state values at the starting point were the same that it was trained upon. However, the gain scaling link between the steady-state model and the dynamic model allow the gain to be scaled and the parameter bi to be scaled such that the dynamic operation is scaled and a more accurate prediction is made which accounts for the dynamic properties of the system.

[0066] Referring now to FIG. 4, there is illustrated a block diagram of a method for determining the parameters ai, bi. This is usually achieved through the use of an identification algorithm, which is conventional. This utilizes the (u(t),y(t)) pairs to obtain the ai and bi parameters. In the preferred embodiment, a recursive identification method is utilized where the ai and bi parameters are updated with each new (ui(t),yi(t)) pair. See: T. Eykhoff, “System Identification”, John Wiley & Sons, New York, 1974, Pages 38 and 39, et. seq., and H. Kurz and W. Godecke, “Digital Parameter-Adaptive Control Processes with Unknown Dead Time”, Automatica, Vol. 17, No. 1, 1981, pp. 245-252, which references are incorporated herein by reference.

[0067] In the technique of FIG. 4, the dynamic model 22 has the output thereof input to a parameter-adaptive control algorithm block 60 which adjusts the parameters in the coefficient storage block 38, which also receives the scaled values of k, bi. This is a system that is updated on a periodic basis, as defined by timing block 62. The control algorithm 60 utilizes both the input u and the output y for the purpose of determining and updating the parameters in the storage area 38.

[0068] Referring now to FIG. 5, there is illustrated a block diagram of the preferred method. The program is initiated in a block 68 and then proceeds to a function block 70 to update the parameters ai, bi utilizing the (u(I),y(I)) pairs. Once these are updated, the program flows to a function block 72 wherein the steady-state gain factor K is received, and then to a function block 74 to set the dynamic gain to the steady state gain, i.e., provide the scaling function described hereinabove. This is performed after the update. This procedure can be used for on-line identification, non-linear dynamic model prediction and adaptive control.

[0069] Referring now to FIG. 6, there is illustrated a block diagram of one application of the present invention utilizing a control environment. A plant 78 is provided which receives input values u(t) and outputs an output vector y(t). The plant 78 also has measurable state variables s(t). A predictive model 80 is provided which receives the input values u(t) and the state variables s(t) in addition to the output value y(t). The steady-state model 80 is operable to output a predicted value of both y(t) and also of a future input value u(t+1). This constitutes a steady-state portion of the system. The predicted steady-state input value is Uss with the predicted steady-state output value being Yss. In a conventional control scenario, the steady-state model 80 would receive as an external input a desired value of the output yd(t) which is the desired value that the overall control system seeks to achieve. This is achieved by controlling a distributed control system (DCS) 86 to produce a desired input to the plant. This is referred to as u(t+1), a future value. Without considering the dynamic response, the predictive model 80, a steady-state model, will provide the steady-state values. However, when a change is desired, this change will effectively be viewed as a “step response”.

[0070] To facilitate the dynamic control aspect, a dynamic controller 82 is provided which is operable to receive the input u(t), the output value y(t) and also the steady-state values Uss and Yss and generate the output u(t+1). The dynamic controller effectively generates the dynamic response between the changes, i.e., when the steady-state value changes from an initial steady-state value Ussi, Yssi to a final steady-state value Ufss, Yfss.

[0071] During the operation of the system, the dynamic controller 82 is operable in accordance with the embodiment of FIG. 2 to update the dynamic parameters of the dynamic controller 82 in a block 88 with a gain link block 90, which utilizes the value Kss from a steady-state parameter block in order to scale the parameters utilized by the dynamic controller 82, again in accordance with the above described method. In this manner, the control function can be realized. In addition, the dynamic controller 82 has the operation thereof optimized such that the path traveled between the initial and final steady-state values is achieved with the use of the optimizer 83 in view of optimizer constraints in a block 85. In general, the predicted model (steady-state model) 80 provides a control network function that is operable to predict the future input values. Without the dynamic controller 82, this is a conventional control network which is generally described in U.S. Pat. No. 5,353,207, issued Oct. 4, 1994, to the present assignee, which patent is incorporated herein by reference.

[0072] Approximate Systematic Modeling

[0073] For the modeling techniques described thus far, consistency between the steady-state and dynamic models is maintained by rescaling the bi parameters at each time step utilizing equation 13. If the systematic model is to be utilized in a Model Predictive Control (MPC) algorithm, maintaining consistency may be computationally expensive. These types of algorithms are described in CE. Garcia, D. M. Prett and M. Morari. Model predictive control: theory and practice—a survey, Automatica, 25:335-348, 1989; D. E. Seborg, T. F. Edgar, and D. A. Mellichamp. Process Dynamics and Control. John Wiley and Sons, New York, N.Y., 1989. These references are incorporated herein by reference. For example, if the dynamic gain kd is computed from a neural network steady-state model, it would be necessary to execute the neural network module each time the model was iterated in the MPC algorithm. Due to the potentially large number of model iterations for certain MPC problems, it could be computationally expensive to maintain a consistent model. In this case, it would be better to use an approximate model which does not rely on enforcing consistencies at each iteration of the model.

[0074] Referring now to FIG. 7, there is illustrated a diagram for a change between steady state values. As illustrated, the steady-state model will make a change from a steady-state value at a line 100 to a steady-state value at a line 102. A transition between the two steady-state values can result in unknown settings. The only way to insure that the settings for the dynamic model between the two steady-state values, an initial steady-state value Kssi and a final steady-state gain Kssf, would be to utilize a step operation, wherein the dynamic gain kd was adjusted at multiple positions during the change. However, this may be computationally expensive. As will be described hereinbelow, an approximation algorithm is utilized for approximating the dynamic behavior between the two steady-state values utilizing a quadratic relationship. This is defined as a behavior line 104, which is disposed between an envelope 106, which behavior line 104 will be described hereinbelow.

[0075] Referring now to FIG. 8, there is illustrated a diagrammatic view of the system undergoing numerous changes in steady-state value as represented by a stepped line 108. The stepped line 108 is seen to vary from a first steady-state value at a level 110 to a value at a level 112 and then down to a value at a level 114, up to a value at a level 116 and then down to a final value at a level 118. Each of these transitions can result in unknown states. With the approximation algorithm that will be described hereinbelow, it can be seen that, when a transition is made from level 110 to level 112, an approximation curve for the dynamic behavior 120 is provided. When making a transition from level 114 to level 116, an approximation gain curve 124 is provided to approximate the steady state gains between the two levels 114 and 116. For making the transition from level 116 to level 118, an approximation gain curve 126 for the steady-state gain is provided. It can therefore be seen that the approximation curves 120-126 account for transitions between steady-state values that are determined by the network, it being noted that these are approximations which primarily maintain the steady-state gain within some type of error envelope, the envelope 106 in FIG. 7.

[0076] The approximation is provided by the block 41 noted in FIG. 2 and can be designed upon a number of criteria, depending upon the problem that it will be utilized to solve. The system in the preferred embodiment, which is only one example, is designed to satisfy the following criteria:

[0077] 1. Computational Complexity: The approximate systematic model will be used in a Model Predictive Control algorithm, therefore, it is required to have low computational complexity.

[0078] 2. Localized Accuracy: The steady-state model is accurate in localized regions. These regions represent the steady-state operating regimes of the process. The steady-state model is significantly less accurate outside these localized regions.

[0079] 3. Final Steady-State: Given a steady-state set point change, an optimization algorithm which uses the steady-state model will be used to compute the steady-state inputs required to achieve the set point. Because of item 2, it is assumed that the initial and final steady-states associated with a set-point change are located in regions accurately modeled by the steady-state model.

[0080] Given the noted criteria, an approximate systematic model can be constructed by enforcing consistency of the steady-state and dynamic model at the initial and final steady-state associated with a set point change and utilizing a linear approximation at points in between the two steady-states. This approximation guarantees that the model is accurate in regions where the steady-state model is well known and utilizes a linear approximation in regions where the steady-state model is known to be less accurate. In addition, the resulting model has low computational complexity. For purposes of this proof, Equation 13 is modified as follows: 7$\begin{array}{cc}{b}_{i,\mathrm{scaled}}=\frac{b_i{K}_{\mathrm{ss}}\left(u\left(t-d-1\right)\right)\left(1+\sum _{i=1}^n{a}_i\right)}{\sum _{i=1}^n{b}_i}& \left(14\right)\end{array}$

[0081] This new equation 14 utilizes Kss(u(t−d−1)) instead of Kss(u(t)) as the consistent gain, resulting in a systematic model which is delay invariant.

[0082] The approximate systematic model is based upon utilizing the gains associated with the initial and final steady-state values of a set-point change. The initial steady-state gain is denoted Kiss while the initial steady-state input is given by Uiss. The final steady-state gain is Kfss and the final input is Ufss. Given these values, a linear approximation to the gain is given by: 8$\begin{array}{cc}{K}_{\mathrm{ss}}\left(u\left(t\right)\right)=K_{\mathrm{ss}}^i+\frac{K_{\mathrm{ss}}^f-K_{\mathrm{ss}}^i}{U_{\mathrm{ss}}^f-U_{\mathrm{ss}}^i}\left(u\left(t\right)-U_{\mathrm{ss}}^i\right).& \left(15\right)\end{array}$

[0083] Substituting this approximation into Equation 13 and replacing u(t−d−1)−ui by δu(t−d−1) yields: 9$\begin{array}{cc}{\tilde{b}}_{j,\mathrm{scaled}}=\frac{b_j{K}_{\mathrm{ss}}^i\left(1+\sum _{i=1}^n{a}_i\right)}{\sum _{i=1}^n{b}_i}+\frac{1}{2}\frac{b_j\left(1+\sum _{i=1}^n{a}_i\right)\left(K_{\mathrm{ss}}^f-K_{\mathrm{ss}}^i\right)}{\left(\sum _{i=1}^n{b}_i\right)\left(U_{\mathrm{ss}}^f-U_{\mathrm{ss}}^i\right)}\delta u\left(t-d-i\right).& \left(16\right)\end{array}$

[0084] To simplify the expression, define the variable bj-Bar as: 10$\begin{array}{cc}{\stackrel{\_}{b}}_j=\frac{b_j{K}_{\mathrm{ss}}^i\left(1+\sum _{i=1}^n{a}_i\right)}{\sum _{i=1}^n{b}_i}& \left(17\right)\end{array}$

[0085] and gj as: 11$\begin{array}{cc}{g}_j=\frac{b_j\left(1+\sum _{i=1}^n{a}_i\right)\left(K_{\mathrm{ss}}^f-K_{\mathrm{ss}}^i\right)}{\left(\sum _{i=1}^n{b}_i\right)\left(U_{\mathrm{ss}}^f-U_{\mathrm{ss}}^i\right)}& \left(18\right)\end{array}$

[0086] Equation 16 may be written as:

{tilde over (b)} j,scaled ={overscore (b)} j +g j δu (t−d−i)  (19)

[0087] Finally, substituting the scaled b's back into the original difference Equation 7, the following expression for the approximate systematic model is obtained: 12$\begin{array}{cc}\delta y\left(t\right)=\sum _{i=1}^n{\stackrel{\_}{b}}_i\delta u\left(t-d-i\right)+\sum _{i=1}^n{g}_i\delta u\left(t-d-i^2\right)\delta u\left(t-d-i\right)-\sum _{i=1}^n{a}_i\delta y\left(t-i\right)& \left(20\right)\end{array}$

[0088] The linear approximation for gain results in a quadratic difference equation for the output. Given Equation 20, the approximate systematic model is shown to be of low computational complexity. It may be used in a MPC algorithm to efficiently compute the required control moves for a transition from one steady-state to another after a set-point change. Note that this applies to the dynamic gain variations between steady-state transitions and not to the actual path values.

[0089] Control System Error Constraints

[0090] Referring now to FIG. 9, there is illustrated a block diagram of the predictin engine for the dynamic controller 82 of FIG. 6. The prediction engine is operable to essentially predict a value of y(t) as the predicted future value y(t+1). Since the prediction engine must determine what the value of the output y(t) is at each future value between two steady-state values, it is necessary to perform these in a “step” manner. Therefore, there will be k steps from a value of zero to a value of N, which value at k=N is the value at the “horizon”, the desired value. This, as will be described hereinbelow, is an iterative process, it being noted that the terminology for “(t+1)” refers to an incremental step, with an incremental step for the dynamic controller being smaller than an incremented step for the steady-state model. For the steady-state model, “y(t+N)” for the dynamic model will be, “y(t+1)” for the steady state The value y(t+1) is defined as follows:

y (t+1)=a1y(t)+a2y(t−1)+b1u(t−d−1)+b2u(t−d−2)  (021)

[0091] With further reference to FIG. 9, the input values u(t) for each (u,y) pair are input to a delay line 140. The output of the delay line provides the input value u(t) delayed by a delay value “d”. There are provided only two operations for multiplication with the coefficients b1 and b2, such that only two values u(t) and u(t−1) are required. These are both delayed and then multiplied by the coefficients b1 and b2 and then input to a summing block 141. Similarly, the output value yp(t) is input to a delay line 142, there being two values required for multiplication with the coefficients a1 and a2. The output of this multiplication is then input to the summing block 141. The input to the delay line 142 is either the actual input value ya(t) or the iterated output value of the summation block 141, which is the previous value computed by the dynamic controller 82. Therefore, the summing block 141 will output the predicted value y(t+1) which will then be input to a multiplexor 144. The multiplexor 144 is operable to select the actual output ya(t) on the first operation and, thereafter, select the output of the summing block 141. Therefore, for a step value of k=0 the value ya(t) will be selected by the multiplexor 144 and will be latched in a latch 145. The latch 145 will provide the predicted value yp(t+k) on an output 146. This is the predicted value of y(t) for a given k that is input back to the input of delay line 142 for multiplication with the coefficients a1 and a2. This is iterated for each value of k from k=0 to k=N.

[0092] The a1 and a2 values are fixed, as described above, with the b1 and b2 values scaled. This scaling operation is performed by the coefficient modification block 38. However, this only defines the beginning steady-state value and the final steady-state value, with the dynamic controller and the optimization routines described in the present application defining how the dynamic controller operates between the steady-state values and also what the gain of the dynamic controller is. The gain specifically is what determines the modification operation performed by the coefficient modification block 38.

[0093] In FIG. 9, the coefficients in the coefficient modification block 38 are modified as described hereinabove with the information that is derived from the steady-state model. The steady-state model is operated in a control application, and is comprised in part of a forward steady-state model 141 which is operable to receive the steady-state input value Uss(t) and predict the steady-state output value Yss(t). This predicted value is utilized in an inverse steady-state model 143 to receive the desired value yd(t) and the predicted output of the steady-state model 141 and predict a future steady-state input value or manipulated value Uss(t+N) and also a future steady-state input value Yss(t+N) in addition to providing the steady-state gain Kss. As described hereinabove, these are utilized to generate scaled b-values. These b-values are utilized to define the gain kd of the dynamic model. In can therefore be seen that this essentially takes a linear dynamic model with a fixed gain and allows it to have a gain thereof modified by a non-linear model as the operating point is moved through the output space.

[0094] Referring now to FIG. 10, there is illustrated a block diagram of the dynamic controller and optimizer. The dynamic controller includes a dynamic model 149 which basically defines the predicted value yp(k) as a function of the inputs y(t), s(t) and u(t). This was essentially the same model that was described hereinabove with reference to FIG. 9. The model 149 predicts the output values yp(k) between the two steady-state values, as will be described hereinbelow. The model 149 is predefined and utilizes an identification algorithm to identify the a1, a2, b1 and b2 coefficients during training. Once these are identified in a training and identification procedure, these are “fixed”. However, as described hereinabove, the gain of the dynamic model is modified by scaling the coefficients b1 and b2. This gain scaling is not described with respect to the optimization operation of FIG. 10, although it can be incorporated in the optimization operation.

[0095] The output of model 149 is input to the negative input of a summing block 150. Summing block 150 sums the predicted output yp(k) with the desired output yd(t). In effect, the desired value of yd(t) is effectively the desired steady-state value Yfss, although it can be any desired value. The output of the summing block 150 comprises an error value which is essentially the difference between the desired value yd(t) and the predicted value yp(k). The error value is modified by an error modification block 151, as will be described hereinbelow, in accordance with error modification parameters in a block 152. The modified error value is then input to an inverse model 153, which basically performs an optimization routine to predict a change in the input value u(t). In effect, the optimizer 153 is utilized in conjunction with the model 149 to minimize the error output by summing block 150. Any optimization function can be utilized, such as a Monte Carlo procedure. However, in the present invention, a gradient calculation is utilized. In the gradient method, the gradient ∂(y)/∂(u) is calculated and then a gradient solution performed as follows: 13$\begin{array}{cc}\Delta u_{\mathrm{new}}=\Delta u_{\mathrm{old}}+\left(\frac{\partial \left(y\right)}{\partial \left(u\right)}\right)\times E& \left(022\right)\end{array}$

[0096] The optimization function is performed by the inverse model 153 in accordance with optimization constraints in a block 154. An iteration procedure is performed with an iterate block 155 which is operable to perform an iteration with the combination of the inverse model 153 and the predictive model 149 and output on an output line 156 the future value u(t+k+1). For k=0, this will be the initial steady-state value and for k=N, this will be the value at the horizon, or at the next steady-state value. During the iteration procedure, the previous value of u(t+k) has the change value Au added thereto. This value is utilized for that value of k until the error is within the appropriate levels. Once it is at the appropriate level, the next u(t+k) is input to the model 149 and the value thereof optimized with the iterate block 155. Once the iteration procedure is done, it is latched. As will be described hereinbelow, this is a combination of modifying the error such that the actual error output by the block 150 is not utilized by the optimizer 153 but, rather, a modified error is utilized. Alternatively, different optimization constraints can be utilized, which are generated by the block 154, these being described hereinbelow.

[0097] Referring now to FIGS. 11a and 11b, there are illustrated plots of the output y(t+k) and the input uk(t+k+1), for each k from the initial steady-state value to the horizon steady-state value at k=N. With specific reference to FIG. 11a, it can be seen that the optimization procedure is performed utilizing multiple passes. In the first pass, the actual value ua(t+k) for each k is utilized to determine the values of y(t+k) for each u,y pair. This is then accumulated and the values processed through the inverse model 153 and the iterate block 155 to minimize the error. This generates a new set of inputs uk(t+k+1) illustrated in FIG. 11b. Therefore, the optimization after pass 1 generates the values of u(t+k+1) for the second pass. In the second pass, the values are again optimized in accordance with the various constraints to again generate another set of values for u(t+k+1). This continues until the overall objective function is reached. This objective function is a combination of the operations as a function of the error and the operations as a function of the constraints, wherein the optimization constraints may control the overall operation of the inverse model 153 or the error modification parameters in block 152 may control the overall operation. Each of the optimization constraints will be described in more detail hereinbelow.

[0098] Referring now to FIG. 12, there is illustrated a plot of yd(t) and yp(t). The predicted value is represented by a waveform 170 and the desired output is represented by a waveform 172, both plotted over the horizon between an initial steady-state value Yiss and a final steady-state value Yfss. It can be seen that the desired waveform prior to k=0 is substantially equal to the predicted output. At k=0, the desired output waveform 172 raises its level, thus creating an error. It can be seen that at k=0, the error is large and the system then must adjust the manipulated variables to minimize the error and force the predicted value to the desired value. The objective function for the calculation of error is of the form: 14$\begin{array}{cc}\underset{\Delta u_{\mathrm{il}}}{\min }\sum _j\sum _k(A_j*{\left({\overrightarrow{y}}^p\left(t\right)-{\overrightarrow{y}}^d\left(t\right)\right)}^2& \left(23\right)\end{array}$

[0099] where:

[0100] Duil is the change in input variable (IV) I at time interval 1

[0101] Aj is the weight factor for control variable (CV) j

[0102] yp(t) is the predicted value of CV j at time interval k

[0103] yd(t) is the desired value of CV j.

[0104] Trajectory Weighting

[0105] The present system utilizes what is referred to as “trajectory weighting” which encompasses the concept that one does not put a constant degree of importance on the future predicted process behavior matching the desired behavior at every future time set, i.e., at low k-values. One approach could be that one is more tolerant of error in the near term (low k-values) than farther into the future (high k-values). The basis for this logic is that the final desired behavior is more important than the path taken to arrive at the desired behavior, otherwise the path traversed would be a step function. This is illustrated in FIG. 13 wherein three possible predicted behaviors are illustrated, one represented by a curve 174 which is acceptable, one is represented by a different curve 176, which is also acceptable and one represented by a curve 178, which is unacceptable since it goes above the desired level on curve 172. Curves 174-178 define the desired behavior over the horizon for k=1 to N.

[0106] In Equation 23, the predicted curves 174-178 would be achieved by forcing the weighting factors Aj to be time varying. This is illustrated in FIG. 14. In FIG. 14, the weighting factor A as a function of time is shown to have an increasing value as time and the value of k increases. This results in the errors at the beginning of the horizon (low k-values) being weighted much less than the errors at the end of the horizon (high k-values). The result is more significant than merely redistributing the weights out to the end of the control horizon at k=N. This method also adds robustness, or the ability to handle a mismatch between the process and the prediction model. Since the largest error is usually experienced at the beginning of the horizon, the largest changes in the independent variables will also occur at this point. If there is a mismatch between the process and the prediction (model error), these initial moves will be large and somewhat incorrect, which can cause poor performance and eventually instability. By utilizing the trajectory weighting method, the errors at the beginning of the horizon are weighted less, resulting in smaller changes in the independent variables and, thus, more robustness.

[0107] Error Constraints

[0108] Referring now to FIG. 15, there are illustrated constraints that can be placed upon the error. There is illustrated a predicted curve 180 and a desired curve 182, desired curve 182 essentially being a flat line. It is desirable for the error between curve 180 and 182 to be minimized. Whenever a transient occurs at t=0, changes of some sort will be required. It can be seen that prior to t=0, curve 182 and 180 are substantially the same, there being very little error between the two. However, after some type of transition, the error will increase. If a rigid solution were utilized, the system would immediately respond to this large error and attempt to reduce it in as short a time as possible. However, a constraint frustum boundary 184 is provided which allows the error to be large at t=0 and reduces it to a minimum level at a point 186. At point 186, this is the minimum error, which can be set to zero or to a non-zero value, corresponding to the noise level of the output variable to be controlled. This therefore encompasses the same concepts as the trajectory weighting method in that final future behavior is considered more important that near term behavior. The ever shrinking minimum and/or maximum bounds converge from a slack position at t=0 to the actual final desired behavior at a point 186 in the constraint frustum method.

[0109] The difference between constraint frustums and trajectory weighting is that constraint frustums are an absolute limit (hard constraint) where any behavior satisfying the limit is just as acceptable as any other behavior that also satisfies the limit. Trajectory weighting is a method where differing behaviors have graduated importance in time. It can be seen that the constraints provided by the technique of FIG. 15 requires that the value yp(t) is prevented from exceeding the constraint value. Therefore, if the difference between yd(t) and yp(t) is greater than that defined by the constraint boundary, then the optimization routine will force the input values to a value that will result in the error being less than the constraint value. In effect, this is a “clamp” on the difference between yp(t) and yd(t). In the trajectory weighting method, there is no “clamp” on the difference therebetween; rather, there is merely an attenuation factor placed on the error before input to the optimization network.

[0110] Trajectory weighting can be compared with other methods, there being two methods that will be described herein, the dynamic matrix control (DMC) algorithm and the identification and command (IdCom) algorithm. The DMC algorithm utilizes an optimization to solve the control problem by minimizing the objective function: 15$\begin{array}{cc}\underset{\Delta U_{\mathrm{il}}}{\min }\sum _j\sum _k(A_j*\left({\overrightarrow{y}}^P\left(t\right)-{\overrightarrow{y}}^D\left(t\right)\right)+\sum _i{B}_i*\sum _1{\left(\Delta U_{\mathrm{il}}\right)}^2& \left(24\right)\end{array}$

[0111] where Bi is the move suppression factor for input variable I. This is described in Cutler, C. R. and B. L. Ramaker, Dynamic Matrix Control—A Computer Control Algorithm, AIChE National Meeting, Houston, Tex. (April, 1979), which is incorporated herein by reference.

[0112] It is noted that the weights Aj and desired values yd(t) are constant for each of the control variables. As can be seen from Equation 24, the optimization is a trade off between minimizing errors between the control variables and their desired values and minimizing the changes in the independent variables. Without the move suppression term, the independent variable changes resulting from the set point changes would be quite large due to the sudden and immediate error between the predicted and desired values. Move suppression limits the independent variable changes, but for all circumstances, not just the initial errors.

[0113] The IdCom algorithm utilizes a different approach. Instead of a constant desired value, a path is defined for the control variables to take from the current value to the desired value. This is illustrated in FIG. 16. This path is a more gradual transition from one operation point to the next. Nevertheless, it is still a rigidly defined path that must be met. The objective function for this algorithm takes the form: 16$\begin{array}{cc}\underset{\Delta U_{\mathrm{il}}}{\min }\sum _j\sum _k{\left(A_j*\left(Y^{P_{\mathrm{jk}}}-y_{\mathrm{refjk}}\right)\right)}^2& \left(25\right)\end{array}$

[0114] This technique is described in Richalet, J., A. Rault, J. L. Testud, and J. Papon, Model Predictive Heuristic Control: Applications to Industrial Processes, Automatica, 14, 413-428 (1978), which is incorporated herein by reference. It should be noted that the requirement of Equation 25 at each time interval is sometimes difficult. In fact, for control variables that behave similarly, this can result in quite erratic independent variable changes due to the control algorithm attempting to endlessly meet the desired path exactly.

[0115] Control algorithms such as the DMC algorithm that utilize a form of matrix inversion in the control calculation, cannot handle control variable hard constraints directly. They must treat them separately, usually in the form of a steady-state linear program. Because this is done as a steady-state problem, the constraints are time invariant by definition. Moreover, since the constraints are not part of a control calculation, there is no protection against the controller violating the hard constraints in the transient while satisfying them at steady-state.

[0116] With further reference to FIG. 15, the boundaries at the end of the envelope can be defined as described hereinbelow. One technique described in the prior art, W. Edwards Deming, “Out of the Crisis,” Massachusetts Institute of Technology, Center for Advanced Engineering Study, Cambridge Mass., Fifth Printing, September 1988, pages 327-329, describes various Monte Carlo experiments that set forth the premise that any control actions taken to correct for common process variation actually may have a negative impact, which action may work to increase variability rather than the desired effect of reducing variation of the controlled processes. Given that any process has an inherent accuracy, there should be no basis to make a change based on a difference that lies within the accuracy limits of the system utilized to control it. At present, commercial controllers fail to recognize the fact that changes are undesirable, and continually adjust the process, treating all deviation from target, no matter how small, as a special cause deserving of control actions, i.e., they respond to even minimal changes. Over adjustment of the manipulated variables therefore will result, and increase undesirable process variation. By placing limits on the error with the present filtering algorithms described herein, only controller actions that are proven to be necessary are allowed, and thus, the process can settle into a reduced variation free from unmerited controller disturbances. The following discussion will deal with one technique for doing this, this being based on statistical parameters.

[0117] Filters can be created that prevent model-based controllers from taking any action in the case where the difference between the controlled variable measurement and the desired target value are not significant. The significance level is defined by the accuracy of the model upon which the controller is statistically based. This accuracy is determined as a function of the standard deviation of the error and a predetermined confidence level. The confidence level is based upon the accuracy of the training. Since most training sets for a neural network-based model will have “holes” therein, this will result in inaccuracies within the mapped space. Since a neural network is an empirical model, it is only as accurate as the training data set. Even though the model may not have been trained upon a given set of inputs, it will extrapolate the output and predict a value given a set of inputs, even though these inputs are mapped across a space that is questionable. In these areas, the confidence level in the predicted output is relatively low. This is described in detail in U.S. patent application Ser. No. 08/025,184, filed Mar. 2, 1993, which is incorporated herein by reference.

[0118] Referring now to FIG. 17, there is illustrated a flowchart depicting the statistical method for generating the filter and defining the end point 186 in FIG. 15. The flowchart is initiated at a start block 200 and then proceeds to a function block 202, wherein the control values u(t+1) are calculated. However, prior to acquiring these control values, the filtering operation must be a processed. The program will flow to a function block 204 to determine the accuracy of the controller. This is done off-line by analyzing the model predicted values compared to the actual values, and calculating the standard deviation of the error in areas where the target is undisturbed. The model accuracy of em(t) is defined as follows:

e m (t)=a(t)−p(t)  (026)

[0119] where:

[0120] em=model error,

[0121] a=actual value

[0122] p=model predicted value

[0123] The model accuracy is defined by the following equation:

Acc=H*σ m   (027)

[0124] where:

[0125] Acc=accuracy in terms of minimal detector error 17$H=\mathrm{significance}\mathrm{level}=167\%\mathrm{confidence}=295\%\mathrm{confidence}=399.5\%\mathrm{confidence}$

[0126] σm=standard deviation of em(t).

[0127] The program then flows to a function block 206 to compare the controller error ec(t) with the model accuracy. This is done by taking the difference between the predicted value (measured value) and the desired value. This is the controller error calculation as follows:

e c (t)=d(t)−m(t)  (028)

[0128] where:

[0129] ec=controller error

[0130] d=desired value

[0131] m=measured value

[0132] The program will then flow to a decision block 208 to determine if the error is within the accuracy limits. The determination as to whether the error is within the accuracy limits is done utilizing Shewhart limits. With this type of limit and this type of filter, a determination is made as to whether the controller error ec(t) meets the following conditions: ec(t)≧−1*Acc and ec(t)≦+1*Ace, then either the control action is suppressed or not suppressed. If it is within the accuracy limits, then the control action is suppressed and the program flows along a “Y” path. If not, the program will flow along the “N” path to function block 210 to accept the u(t+1) values. If the error lies within the controller accuracy, then the program flows along the “Y” path from decision block 208 to a function block 212 to calculate the running accumulation of errors. This is formed utilizing a CUSUM approach. The controller CUSUM calculations are done as follows:

S low =min(0,Slow(t−1)+d(t)−m(t))−Σ(m)+k)  (029)

S hi =max(0,Shi(t−1)+[d(t)−m(t))−Σ(m)]−k)  (030)

[0133] where:

[0134] Shi=Running Positive Qsum

[0135] Slow=Running Negative Qsum

[0136] k=Tuning factor—minimal detectable change threshold

[0137] with the following defined:

[0138] Hq=significance level. Values of (j,k) can be found so that the CUSUM control chart will have significance levels equivalent to Shewhart control charts.

[0139] The program will then flow to a decision block 214 to determine if the CUSUM limits check out, i.e., it will determine if the Qsum values are within the limits. If the Qsum, the accumulated sum error, is within the established limits, the program will then flow along the “Y” path. And, if it is not within the limits, it will flow along the “N” path to accept the controller values u(t+1). The limits are determined if both the value of Shi≧+1*Hq and Slow≦−1*Hq. Both of these actions will result in this program flowing along the “Y” path. If it flows along the “N” path, the sum is set equal to zero and then the program flows to the function block 210. If the Qsum values are within the limits, it flows along the “Y” path to a function block 218 wherein a determination is made as to whether the user wishes to perturb the process. If so, the program will flow along the “Y” path to the function block 210 to accept the control values u(t+1). If not, the program will flow along the “N” path from decision block 218 to a function block 222 to suppress the controller values u(t+1). The decision block 218, when it flows along the “Y” path, is a process that allows the user to re-identify the model for on-line adaptation, i.e., retrain the model. This is for the purpose of data collection and once the data has been collected, the system is then reactivated.

[0140] Referring now to FIG. 18, there is illustrated a block diagram of the overall optimization procedure. In the first step of the procedure, the initial steady-state values {Yssi, Ussi} and the final steady-state values {Yssf, Ussf} are determined, as defined in blocks 226 and 228, respectively. In some calculations, both the initial and the final steady-state values are required. The initial steady-state values are utilized to define the coefficients ai, bi in a block 228. As described above, this utilizes the coefficient scaling of the b-coefficients. Similarly, the steady-state values in block 228 are utilized to define the coefficients af, bf, it being noted that only the b-coefficients are also defined in a block 229. Once the beginning and end points are defined, it is then necessary to determine the path therebetween. This is provided by block 230 for path optimization. There are two methods for determining how the dynamic controller traverses this path. The first, as described above, is to define the approximate dynamic gain over the path from the initial gain to the final gain. As noted above, this can incur some instabilities. The second method is to define the input values over the horizon from the initial value to the final value such that the desired value Yssf is achieved. Thereafter, the gain can be set for the dynamic model by scaling the b-coefficients. As noted above, this second method does not necessarily force the predicted value of the output yp(t) along a defined path; rather, it defines the characteristics of the model as a function of the error between the predicted and actual values over the horizon from the initial value to the final or desired value. This effectively defines the input values for each point on the trajectory or, alternatively, the dynamic gain along the trajectory.

[0141] Referring now to FIG. 18a, there is illustrated a diagrammatic representation of the manner in which the path is mapped through the input and output space. The steady-state model is operable to predict both the output steady-state value Yssi at a value of k=0, the initial steady-state value, and the output steady-state value Yssi at a time t+N where k=N, the final steady-state value. At the initial steady-state value, there is defined a region 227, which region 227 comprises a surface in the output space in the proximity of the initial steady-state value, which initial steady-state value also lies in the output space. This defines the range over which the dynamic controller can operate and the range over which it is valid. At the final steady-state value, if the gain were not changed, the dynamic model would not be valid. However, by utilizing the steady-state model to calculate the steady-state gain at the final steady-state value and then force the gain of the dynamic model to equal that of the steady-state model, the dynamic model then becomes valid over a region 229, proximate the final steady-state value. This is at a value of k=N. The problem that arises is how to define the path between the initial and final steady-state values. One possibility, as mentioned hereinabove, is to utilize the steady-state model to calculate the steady-state gain at multiple points along the path between the initial steady-state value and the final steady-state value and then define the dynamic gain at those points. This could be utilized in an optimization routine, which could require a large number of calculations. If the computational ability were there, this would provide a continuous calculation for the dynamic gain along the path traversed between the initial steady-state value and the final steady-state value utilizing the steady-state gain. However, it is possible that the steady-state model is not valid in regions between the initial and final steady-state values, i.e., there is a low confidence level due to the fact that the training in those regions may not be adequate to define the model therein. Therefore, the dynamic gain is approximated in these regions, the primary goal being to have some adjustment of the dynamic model along the path between the initial and the final steady-state values during the optimization procedure. This allows the dynamic operation of the model to be defined. This is represented by a number of surfaces 225 as shown in phantom.

[0142] Referring now to FIG. 19, there is illustrated a flow chart depicting the optimization algorithm. The program is initiated at a start block 232 and then proceeds to a function block 234 to define the actual input values ua(t) at the beginning of the horizon, this typically being the steady-state value Uss. The program then flows to a function block 235 to generate the predicted values yp(k) over the horizon for all k for the fixed input values. The program then flows to a function block 236 to generate the error E(k) over the horizon for all k for the previously generated yp(k). These errors and the predicted values are then accumulated, as noted by function block 238. The program then flows to a function block 240 to optimize the value of u(t) for each value of k in one embodiment. This will result in k-values for u(t). Of course, it is sufficient to utilize less calculations than the total k-calculations over the horizon to provide for a more efficient algorithm. The results of this optimization will provide the predicted change Δu(t+k) for each value of k in a function block 242. The program then flows to a function block 243 wherein the value of u(t+k) for each u will be incremented by the value Δu(t+k). The program will then flow to a decision block 244 to determine if the objective function noted above is less than or equal to a desired value. If not, the program will flow back along an “N” path to the input of function block 235 to again make another pass. This operation was described above with respect to FIGS. 11a and 11b. When the objective function is in an acceptable level, the program will flow from decision block 244 along the “Y” path to a function block 245 to set the value of u(t+k) for all u. This defines the path. The program then flows to an End block 246.

[0143] Steady State Gain Determination

[0144] Referring now to FIG. 20, there is illustrated a plot of the input space and the error associated therewith. The input space is comprised of two variables x1 and x2. The y-axis represents the function f(x1, x2). In the plane of x1 and x2, there is illustrated a region 250, which represents the training data set. Areas outside of the region 250 constitute regions of no data, i.e., a low confidence level region. The function Y will have an error associated therewith. This is represented by a plane 252. However, the error in the plane 250 is only valid in a region 254, which corresponds to the region 250. Areas outside of region 254 on plane 252 have an unknown error associated therewith. As a result, whenever the network is operated outside of the region 250 with the error region 254, the confidence level in the network is low. Of course, the confidence level will not abruptly change once outside of the known data regions but, rather, decreases as the distance from the known data in the training set increases. This is represented in FIG. 21 wherein the confidence is defined as α(x). It can be seen from FIG. 21 that the confidence level α(x) is high in regions overlying the region 250.

[0145] Once the system is operating outside of the training data regions, i.e., in a low confidence region, the accuracy of the neural net is relatively low. In accordance with one aspect of the preferred embodiment, a first principles model g(x) is utilized to govern steady-state operation. The switching between the neural network model f(x) and the first principle models g(x) is not an abrupt switching but, rather, it is a mixture of the two.

[0146] The steady-state gain relationship is defined in Equation 7 and is set forth in a more simple manner as follows: 18$\begin{array}{cc}K\left(\overrightarrow{u}\right)=\frac{\partial \left(f\left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}& \left(031\right)\end{array}$

[0147] A new output function Y(u) is defined to take into account the confidence factor α(u) as follows:

Y ({right arrow over (u)})=α({right arrow over (u)}).f({right arrow over (u)})+(1−α({right arrow over (u)}))g({right arrow over (u)})  (032)

[0148] where:

[0149] α(u)=confidence in model f(u)

[0150] α(u) in the range of 0→1

[0151] α(u)ε{0,1}

[0152] This will give rise to the relationship: 19$\begin{array}{cc}K\left(\overrightarrow{u}\right)=\frac{\partial \left(Y\left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}& \left(033\right)\end{array}$

[0153] In calculating the steady-state gain in accordance with this Equation utilizing the output relationship Y(u), the following will result: 20$\begin{array}{cc}K\left(\overrightarrow{u}\right)=\frac{\partial \left(\alpha \left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}\times F\left(\overrightarrow{u}\right)+\alpha \left(\overrightarrow{u}\right)\frac{\partial \left(F\left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}+\frac{\partial \left(1-\alpha \left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}\times g\left(\overrightarrow{u}\right)+\left(1-\alpha \left(\overrightarrow{u}\right)\right)\frac{\partial \left(g\left(\overrightarrow{u}\right)\right)}{\partial \left(\overrightarrow{u}\right)}& \left(034\right)\end{array}$

[0154] Referring now to FIG. 22, there is illustrated a block diagram of the embodiment for realizing the switching between the neural network model and the first principles model. A neural network block 300 is provided for the function f(u), a first principle block 302 is provided for the function g(u) and a confidence level block 304 for the function α(u). The input u(t) is input to each of the blocks 300-304. The output of block 304 is processed through a subtraction block 306 to generate the function 1-α(u), which is input to a multiplication block 308 for multiplication with the output of the first principles block 302. This provides the function (1−α(u))*g(u). Additionally, the output of the confidence block 304 is input to a multiplication block 310 for multiplication with the output of the neural network block 300. This provides the function f(u)*α(u). The output of block 308 and the output of block 310 are input to a summation block 312 to provide the output Y(u).

[0155] Referring now to FIG. 23, there is illustrated an alternate embodiment which utilizes discreet switching. The output of the first principles block 302 and the neural network block 300 are provided and are operable to receive the input x(t). The output of the network block 300 and first principles block 302 are input to a switch 320, the switch 320 operable to select either the output of the first principals block 302 or the output of the neural network block 300. The output of the switch 320 provides the output Y(u).

[0156] The switch 320 is controlled by a domain analyzer 322. The domain analyzer 322 is operable to receive the input x(t) and determine whether the domain is one that is within a valid region of the network 300. If not, the switch 320 is controlled to utilize the first principles operation in the first principles block 302. The domain analyzer 322 utilizes the training database 326 to determine the regions in which the training data is valid for the network 300. Alternatively, the domain analyzer 320 could utilize the confidence factor α(u) and compare this with a threshold, below which the first principles model 302 would be utilized.

[0157] Non-Linear Mill Control

[0158] Overall, model predictive control (MPC) has been the standard supervisory control tool for such processes as are required in the cement industry. In the cement industry, particulate is fabricated with a kiln/cooler to generate raw material and then to grind this material with a mill. The overall kiln/cooler application, in the present embodiment, utilizes a model of the process rather than a model of the operator. This model will provide continuous regulation and disturbance rejection which will allow the application to recover from major upsets, such as coating drop three times faster than typical operator intervention.

[0159] In general, mills demonstrate a severe non-linear behavior. This can present a problem in various aspects due to the fact that the gains at different locations within the input space can change. The cement kilns and coolers present a very difficult problem, in that the associated processes, both chemical and physical, are in theory simple, but in practice complex. This is especially so when commercial issues such as quality and costs of production are considered. The manufacturing of cement, and its primary ingredient, clinker, has a number of conflicting control objectives, which are to maximize production, minimize costs, and maximize efficiency, while at the same time maintaining minimum quality specifications. All of this optimization must take place within various environmental, thermodynamic and mechanical constraints.

[0160] A primary technique of control for clinker has been the operator. As rotary cement kilns and automation technology evolve, various automation solutions have been developed for the cement industry. These solutions have been successful to a greater or lessor extent. In the present application, the process is modeled, rather than the operator, and model predictive control is utilized. Moves are made every control cycle to the process based on continuous feedback of key measurements. This gives rise to a continuous MPC action, as opposed to the intermittent, albeit frequent moves made by the typical expert system. In addition, as will be described hereinbelow, the approach described utilizes full multivariable control (MVC) techniques, which take into account all coupled interactions in the kiln/cooler process.

[0161] The cement mill is utilized to manufacture the various grades of cement after processing of the raw material, which are defined by their chemical composition and fineness (particle size distribution). The control objectives are thus to maximize production at minimum cost, i.e., low energy consumption for the various product grades, chemical compositions and specified fineness. In general, the mill utilizes a closed circuit where separators in the feed-back are utilized to classify the mill output into oversized and undersized product streams. The oversized stream, which does not conform to specification required for correct cement strength, is fed back into the mill for further grinding and size reduction. Depending upon the type of mill, controls include fresh feed, recirculating-load, as well as separator speed, all of which are used by the operator to control fineness, energy consumption and throughput.

[0162] In general, the mill grinding equations take the form of:

ln (P)=k1+k2*F

[0163] where:

[0164] P=particle size

[0165] F=feed rate

[0166] k1 and k2 are constants.

[0167] It has generally been stated in the literature that grinding model equations are non-linear and hence, direct application of linear control theory is not possible. The primary reason for this is that the operation of the plant is only non-linear in very small regions of the input space. Once the process has traversed, i.e., “stepped,” from one portion of the input space to another portion thereof, the overall model changes, i.e., it is non-linear. This lends itself to non-linear modeling techniques. However, most control systems are linear in nature, especially those that model the dynamics of the plant.

[0168] Referring now to FIG. 24, there is illustrated a diagrammatic view of the kiln/cooler configuration and the selected instrumentation utilized for optimal MPC control. This kiln/cooler consists of a five-stage suspension pre-heater kiln, with back-end firing (approximately fifteen percent of total firing). The cooler is a grate type with a conversion upgrade on the first section. It has on-line analyzers for NOx, O2 and CO located at the top of a preheater 2402 which receives raw meal therein. The output of the preheater is input to a kiln 2404. There is provided a coal feed on an input 2406, the feed end and a coal feed 2408 on the firing end. The output of the kiln is input to a cooler 2410 which has an input cooler fan 2412. The output of the cooler provides the clinker. The overall plan is fully instrumented with all necessary measurements, e.g., temperature, pressure and flow measurements such as coal, raw meal and grate air. The quality of the clinker production is maintained by the analysis of hourly samples of the raw meal feed, clinker and coal. This is supported by a semi-automated sampling system, and various modem laboratory infrastructure.

[0169] This system is controlled with an MPC controller (not shown) that consists of the MPC control engine as described hereinabove, as well as a real-time expert system that performs a variety of pre and post processing of control signals as well as various other functions such as MPC engine control, noise filtering, bias compensation and real-time trending. This system will also perform set point tracking for bumperless transfer, and adaptive target selection. This allows for the controller tuning parameters to be changed according to various business and/or process strategies.

[0170] The MPC is defined in two primary phases the first being the modeling phase in which the models of the kiln processes are developed. The second phase is the deployment phase, where the models are commissioned and refined to a point where satisfactory control can be obtained. Central to commissioning is the tuning where the controller is tweaked to provide the desired control and optimization. For example this could be: maximum production at the expense of quality, or optimal quality at the expense of efficiency.

[0171] The MPC models are developed from the analysis of test and process data, together with knowledge from the plant operators and other domain experts. The result is a matrix of time responses, where each response reflects the dynamic interaction of a controlled variable to a manipulated variable.

[0172] The tuning involves the selection of targets (setpoints), weighting factors and various constraints for each variable. This determines how the controller will solve the control problem at any given time. The control of the kiln and its optimization within the above set of constraints is solved every control cycle.

[0173] The solution chosen in a particular control cycle may not seem to be necessarily optimal at that given time, but will be optimal within the solution space which has temporal as well as spatial dimensions. Thus the control solution is a series of trajectories into the future, where the whole solution is optimized with time. The very nature of optimal control in real time does not allow for a guarantee of a global optimal solution. However the calculation of an optimal solution within a finite amount of time is itself a class of optimization.

[0174] Some of the tuning parameters, which can be changed during operations, include:

[0175] 1) Targets. Targets can be set for both controlled and manipulated variables, and the MPC controller will try and force all variables to their desired targets. In particular setting a target for a manipulated variable such as coal allows for optimization, and in particular efficiency, because the controller will continually seek a lower coal flow while maintaining production and quality. For some variables such as O2, a target may not be necessary, and the variables will be allowed to roam within a band of constraints.

[0176] 2) Priorities. Setting relative priorities between manipulated variables and controlled variables allows the controller to prioritize which are more important problems to solve, and what type of solution to apply. Underspecified multivariable control (more manipulated variables than controlled variables, as is the case in this application) implies that for every problem there will be more than one solution, but within constraints one solution will generally be more optimal than others. For example, too high a hood temperature can be controlled by, (a) reducing fuel, (b) increasing the grate speed, or (c) increasing cooler airflow, or a combination of the above.

[0177] 3) Hard Constraints. Setting upper and lower hard constraints for each process variable, for example, minimum and maximum grate speed. These values which are usually defined by the mechanical and operational limitations of the grate. Maintaining these constraints is obviously realizable with controlled variables such as ID-fan speed, but is more difficult to achieve with, for example, hood temperature. However when hood temperature exceeds a upper hard constraint of say 1200° C., the controller will switch priority to this temperature excursion, and all other control problems will “take a back seat” to the solution required to bring this temperature back into the allowable operating zone.

[0178] 4) Soft upper and lower constraints. If any process variable penetrates into the soft constraint area, penalties will be incurred that will begin to prioritize the solution of this problem. Continuous penetration into this area will cause increasing prioritization of this problem, thus in effect creating an adaptive prioritization, which changes with the plant state.

[0179] 5) Maximum rate of change constraints. These parameters are only applicable to the manipulated variables, and generally reflect a mechanical of physical limitation of the equipment used, for example maximum coal feed rate.

[0180] From a clinker production point of view the functions of the MPC application can be viewed as follows:

[0181] 1) Kiln Combustion Control where manipulated variables such as ID-fan speed and fuel-flow are manipulated to control primarily O2. When CO rises above a specified threshold constraint, it will override and become the controlled variable. The controller is tuned to heavily penalize high CO values, steer the CO back into an acceptable operating region, and rapidly return to O2 control.

[0182] 2) Kiln Thermal “Hinge Point” Control adjusts total coal, cooler grate speed, and cooler fans to control the hood temperature. The hood temperature is conceptualized as the “hinge” point on which the kiln temperature profile hangs. The controller is tuned to constantly minimize cooler grate speed and cooler fans, so that heat recovery from the cooler is maximized, while minimizing moves to coal.

[0183] 3) Kiln Thermal “Swing Arm” Control adjusts percent coal to the kiln backend, in order to control clinker free lime based on hourly lab feedback. This control function is about three times slower than the hinge point control, which maintains hood temperature at a fixed target. The “swing arm effect” raises or lowers the back end temperature with a constant firing end temperature to compensate for changes in free lime. This is in effect part of the quality control.

[0184] Kiln combustion control, kiln thermal hinge point control, and kiln thermal swing arm control are implemented in a single MPC controller. Kiln speed is included as a disturbance variable, as the production philosophy, in one embodiment, calls for setting a production rate to meet various commercial obligations. This means that any changes to kiln speed and hence production rate by the operator will be taken into account in the MPC predictions, but the MPC controller will not be able to move kiln speed.

[0185] The control system allows for customization of the interface between the plant and the MPC special control functions, the special control functions implemented including:

[0186] 1) Total Coal Control allows the operator to enter a total coal or fuel flow setpoint. The control system “wrapper” splits the move to the front and back individual coal flow controllers while maintaining the percent of coal to the back constant. The purpose of this control function is to allow heating and cooling of the kiln while maintaining a constant energy profile from the preheaters through to the firing end of the kiln. This provides a solid basis for the temperature “hinge point” advanced control function previously described.

[0187] 2) Percent Coal to the Back Control allows the operator to enter a percent coal to the back target and implements the moves to the front and back coal flow controllers to enforce it. The purpose of this control is to allow the back end temperature to be swung up or down by the thermal “swing arm” advanced control function.

[0188] 3) Feed-to-Speed Ratio Control adjusts raw meal feed to the kiln to maintain a constant ratio to kiln speed. The purpose of this controller is to maintain a constant bed depth in the kiln, which is important for long-term stabilization.

[0189] 4) Cooler Fans Control is a move splitter that relates a single generic cooler air fans setpoint to actual setpoints required by n cooler air fans. The expert system wrapper through intelligent diagnostics or by operator selection can determine which of the air cooler fans will be placed under direct control of the MPC controller, thus allowing for full control irrespective of the (for example) maintenance being undertaken on any fans.

[0190] 5) Gas analyzer selection. The control system automatically scans the health of the gas analyzers, and will switch to the alternative analyzer should the signals become “unhealthily’. In addition the control system is used to intelligently extract the fundamental control signals from the O2 and CO readings, which are badly distorted by purge spikes etc.

[0191] Referring now to FIG. 25, there is illustrated a block diagram of the non-linear mill and the basic instrumentation utilized for advanced control therein. The particle size overall is measured as “Blaine” in cm2/gm , and is controlled by the operator through adjustment of the fresh feed rate and the separator speed. The mill is a ball mill, which is referred to by reference numeral 2502. The fresh feed is metered by a fresh feed input device 2506 which receives mined or processed material into the mill 2502, which mill is a mechanical device that grinds this mined or processed material. In this embodiment, the mill is a ball mill, which is a large cylinder that is loaded with steel balls. The mill 2502 rotates and is controlled by a motor 2508 and, as the material passes therethrough, it is comminuted to a specified fineness or Blaine. The output of the mill is input to an elevator 2510 which receives the output of the mill 2502 and inputs it to a separator 2512. This is a feedback system which is different than an “open circuit” mill and is referred to as a “closed circuit” mill. The mill product is fed into the separator 2512, which separator 2512 then divides the product into two streams. The particles that meet product specification are allowed to exit the system as an output, represented by a reference numeral 2514, whereas the particles that are too large are fed back to the input of the mill 2502 through a return 2516 referred to as the course return.

[0192] There are provided various sensors for the operation of the mill. The separator speed is controlled by an input 2518 which signal is generated by a controller 2520. The elevator 2510 provides an output 2522, which constitutes basically the current required by the elevator 2510. This can be correlated to the output of the mill, as the larger the output, the more current that is required to lift it to the separator 2512. Additionally, the motor 2508 can provide an output. There is additionally provided an “ear,” which is a sonic device that monitors the operation of the mill through various sonic techniques. It is known that the operation of the mill can be audibly detected such that operation within certain frequency ranges indicates that the mill is running well and in other frequency ranges that it is not running well, i.e., it is not optimum.

[0193] Overall, the mill-separator-return system is referred to as a “mill circuit.”The main control variable for a mill circuit is product particle size, the output, and fresh feed is manipulated to control this variable. A secondary control variable is return and separator speed is manipulated to control this variable. There are also provided various constants as inputs and constraints for the control operation. This controller 2520 will also control fresh feed on a line 2524.

[0194] The response of particle size to a move in fresh feed is known to be slow (one-two hours) and is dominated by dead time. Where a dead time to time constant ratio exceeding 0.5 is known to be difficult to control without model predictive control techniques, documents ratios for the response of particle size to a move in fresh feed includes 0.9 and 1.3.

[0195] In the case of a closed-circuit mill, a move to fresh feed effects not only the product particle size, but also the return flow. Also, a move to separator speed effects not only the return flow, but also the particle size. This is a fully interactive multi-variable control problem.

[0196] The controller adjusts fresh feed and separator speed to control Blaine and return. It also includes motor and sonic ear as outputs, and the sonic ear is currently used as a constraint variable. That means when the sonic ear decibel reading is too high then fresh feed is decreased. In this way the controller maximizes feed to the sonic ear (choking) constraint.

[0197] Referring now to FIG. 26, there is illustrated a dynamic model matrix for the fresh feed and the separator speed for the measured variables of the Blaine Return Ear and motor. It can be seen that each of these outputs has a minimum and maximum gain associated therewith dead-time delay and various time constants.

[0198] Referring now to FIG. 27, there is illustrated a plot of log sheet data utilized to generate a gain model for the controller 2520. This constitutes the historical data utilized to train the steady state non-linear model.

[0199] In general, the operation described hereinabove utilizes a non-linear controller which provides a model of the dynamics of the plants in a particular region. The only difference in the non-linear model between one region of the input space to a second region of the input space is that associated with the dynamic gain “k.” This dynamic gain varies as the input space is traversed, i.e., the model is only valid over a small region of the input space for a given dynamic gain. In order to compensate for this dynamic gain of a dynamic linear model, i.e., the controller, a non-linear steady state model of the overall process is utilized to calculate a steady-state gain “K” which is then utilized to modify the dynamic gain “k.” This was described in detail hereinabove. In order to utilize this model, it is necessary to first model the non-linear operation of the process, i.e., determining a non-linear steady state model, and then also determine the various dynamics of the system through “step testing.” The historical data provided by the log sheets of FIG. 27 provide this information which can then be utilized to train the non-linear steady state model.

[0200] Referring now to FIGS. 28a and 28b, there are illustrated plots of process gains for the overall non-linear mill model with the scales altered. In FIG. 28a, there is illustrated a plot of the sensitivity versus the percent move of the Blaine. There are provided two plots, one for separator speed and one for fresh feed. The upper plot illustrates that the sensitivity of Blaine to separator speed is very low, whereas the gain of the percent movement of fresh feed with respect to the Blaine varies considerably. It can be seen that at a minimum, the gain is approximately −1.0 and then approaches 0.0 as Blaine increases. In the plot of FIG. 28b, there are illustrated two plots for the sensitivity versus the percent return. In the upper plot, that associated with the percent movement of separator speed, it can be seen that the gain varies from a maximum of 0.4 at a low return to a gain of 0.0 at a high return value. The fresh feed varies similarly, with less range. In general, the plots of FIGS. 28a and 28a, it can be seen that the sensitivity of the control variable, e.g., Blaine in FIG. 28a and return in the plot of FIG. 28b, as compared to the manipulated variables of separator speed and fresh feed. The manipulated variables are illustrated as ranging from their minimum to maximum values.

[0201] In operation of the controller 2520, the process gains (sensitivities) are calculated from the neural network model (steady state model) at each control execution, and downloaded into the predictive models in the linear controller (dynamic model). This essentially provides the linear controller with the dynamic gains that follow the sensitivities exhibited by the steady-state behavior of the process, and thereby provides non-linear mill control utilizing a linear controller. With such a technique, the system can now operate over a much broader specification without changing the tuning parameters or the predictive models in the controller.

[0202] Referring now to FIG. 29, there is illustrated a plot of non-linear gains for fresh feed responses for both manipulatible variables and control variables. The manipulatible variable for the fresh feed is stepped from a value of 120.0 to approximately 70.0. It can be seen that the corresponding Blaine output and the return output, the controlled variables, also steps accordingly. However, it can be seen that a step size of 10.0 in the fresh feed input does not result in identical step in either the Blaine or the return over the subsequent steps. Therefore, it can be seen that this is clearly a non-linear system with varying gains.

[0203] Referring now to FIG. 30, there is illustrated four plots of the non-linear gains for separator speed responses, wherein the separator speed as a manipulatible variable is stepped from a value of approximately 45.0 to 85.0, in steps of approximately 10.0. It can be seen that the return controlled variable makes an initial change with an initial step that is fairly large compared to any change at the end. The Blaine, by comparison, appears to change an identical amount each time. However, it can be seen that the response of the return with respect to changes in the separator speed will result in a very non-linear system.

[0204] Referring now to FIG. 31, there is illustrated an overall output for the closed loop control operation wherein a target change for the Blaine is provided. In this system, it can be seen that the fresh feed is stepped at a time 3102 with a defined trajectory and this results in the Blaine falling and the return rising and falling. Further, the separator speed is also changed from one value to a lower value. However, the manipulatible variables have the trajectory thereof defined by the non-linear controller such that there is an initial point and a final point and a trajectory therebetween. By correctly defining this trajectory between the initial point and the final point, the trajectory of the Blaine and the return can be predicted. This is due to the fact that the non-linear model models the dynamics of the plant, these dynamics being learned from the step tests on the plant. This, of course, is learned at one location in the input space and, by utilizing the steady state gains from the non-linear steady state model to manipulate the dynamic gains of the linear model, control can be effected over different areas of the input space, even though the operation is non-linear from one point in the input space to another space.

[0205] Although the preferred embodiment has been described in detail, it should be understood that various changes, substitutions and alterations can be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A method for controlling a non-linear plant, comprising the steps of: providing a linear controller having a linear gain k that is operable to receive inputs representing measured variables of the plant and predicting on an output of the linear controller predicted control values for manipulatible variables that control the plant; providing a non-linear model of the plant for storing a representation of the plant over a trained region of the operating input space and having a steady-state gain K associated therewith; adjusting the gain k of the linear model with the gain K of the non-linear model in accordance with a predetermined relationship as the measured variables change the operating region of the input space at which the linear controller is predicting the values for the manipulatible variables; and outputting the predictive manipulatible variables after the step of adjusting the gain k.

2. The method of claim 1, wherein the linear controller is operable to model the dynamics of the plant.

3. The method of claim 1, wherein the dynamics of the plant are modeled over a defined region of the input space.

4. The method of claim 1, wherein the training region of the operating input space over which the non-linear model is representative of the operation of the plant represents a space greater than that over which the linear controller is valid.

5. The method of claim 1, and further comprising the step of controlling the operation of the plant with the predicted control variables after the step of adjusting.