WO2002022246A1

WO2002022246A1 - System and method for mixture optimization

Info

Publication number: WO2002022246A1
Application number: PCT/US2001/042151
Authority: WO
Inventors: Juan M. Martin Sanchez
Original assignee: Thermo Gamma-Metrics
Priority date: 2000-09-11
Filing date: 2001-09-11
Publication date: 2002-03-21
Also published as: AU2001295039A1

Abstract

The system and method for mix optimization in a raw mix process combine an on-line analyzer of material composition with a system controller which includes software for processing measurements and controlling flow rates of the raw materials used to make the raw mix. Periodic measurements by the analyzer generate a plurality of quality control parameters that characterize the raw mix composition. The system controller processes the measurements to generate an advanced control signal for each quality control parameter. The advanced control signals are generally defined ratios between functions of the flow rates to be applied to the different sources of the raw materials in the mix. Using the advanced control signals, flow rate set point values for the different sources of raw materials are provided to the local flow rate controllers. When deficiencies in the elemental composition of one or more sources makes it impossible to obtain the quality control parameter target, the system and method generate corresponding diagnoses and can use corrective sources to compensate for the deficiencies.

Description

SYSTEM AND METHOD FOR MIXTURE OPTIMIZATION

FIELD OF THE INVENTION

The invention relates to a system and method for optimizing mixture of raw materials from different sources with different known or unknown elemental compositions.

BACKGROUND OF THE INVENTION

A number of diverse industries rely upon the blending of raw materials to generate their products, which may include cement, ceramics and energy, among others. One example of a process in which optimization can be applied are the raw mix processes of the cement industry, in which the quality control parameters of the mix for the kiln feed are controlled by proportioning the flow rates of different sources of raw materials such as limestone, sandstone and sweetener. In a cement plant, transport belts, e.g. weighfeeders, transport each of the raw materials to a mixer which mixes the materials together. A raw mill receives the mixed material and grinds and blends it into a powder, known as a "raw mix". The raw mill feeds the raw mix to a kiln where it undergoes a calcination process. In order to produce a quality cement, it is necessary that the raw mix produced by the raw mill have physical properties with certain desirable values. Some of the physical properties which may be used to characterize the raw mix are Lime Saturation Factor (LSF), Silica Modulus (SM) and Iron Modulus (IM). These properties, usually called quality control (QC) parameters or modules, are all known functions of four metallic oxides (i.e., calcium, iron, aluminum, and siicon) present in each of the raw materials.

Further considering the cement industry example, until recently, the proportioning of raw materials from different sources in the raw mix process was achieved based on XRF (x-ray fluorescence) laboratory analysis of specimens extracted periodically from the mix, typically at intervals of an hour or more. After receiving the analysis results for the extracted samples, an operator would adjust the proportioning. Using this method, the long period between samples and the time lag between sampling and response severely limited the perfom ance of control systems.

On-line Prompt Gamma Neutron Activation Analyzers (PGNAA), which were introduced to the market in the early- to mid-1980s, are able to measure elemental composition of material in real time at intervals of one minute of less. This high frequency measurement capability opened the door for development of a high performance automatic control system for raw mix processes, however, additional issues needed to be resolved before such process control methods could be realized. For example, raw mix processes using naturally-occurring source materials, e.g., mined ores and quarried materials, are of a multivariable, non-linear nature because the concentrations of the components in the different sources are generally unknown and time-varying. In many cases, these processes also operate with transport time delays.. Therefore, the relationship between the flow rates of the different sources and the quality control parameters of the mix is unknown, time-varying, complex and non-linear. Traditional control solutions have, thus far, been unable to provide a satisfactory solution, and adaptive/predictive solutions have been adopted with mixed results.

Using an adaptive/predictive approach, measurements of elemental composition provided by the analyzer are combined with the source flow rates to estimate the elemental composition of the sources. Using the estimated source composition and the quality control parameter target for the mix, the source flow rates are calculated so that the predicted value of the quality control parameters is as close as possible to the target. However, due to errors in the estimated source compositions, the predicted value rarely matched the target. To compensate for this error, a conventional PID (proportional- integral-derivative) controller was used with a driving error signal provided by the difference between the quality control parameter measurements and the target, and a control signal acting on the quality control parameter target used by the predictive scheme. However, this kind of raw mix optimization process with time-varying and nonlinear characteristics are difficult to control with fixed parameter PID controllers. Accordingly, the need remains for a system and method for optimizing the proportioning of source materials for production of raw mix. SUMMARY OF THE INVENTION

The system and method for mix optimization in a raw mix process combine an on-line analyzer of material composition with a system controller which includes software for processing measurements and controlling flow rates of the raw materials used to make the raw mix. Periodic measurements by the analyzer generate a plurality of quality control parameters that characterize the raw mix composition. Using an advanced control tool such as those known in the art, the system controller processes the measurements to generate an advanced control signal for each quality control parameter. The advanced control signals are generally defined ratios between functions of the flow rates to be applied to the different sources of the raw materials in the mix. Using the advanced control signals, flow rate values for the different sources of raw materials are provided to the local flow rate controllers. When deficiencies in the elemental composition of one or more sources makes it impossible to obtain the quality control parameter target, the system and method generate a corresponding diagnosis.

The system and method for raw mix optimization implements an advanced control tool for real time control of the quality control parameters of the mix product. Raw mix processes obtain a mix composition with a desired quality by combining proportioned quantities of raw materials coming from different sources which have different elemental compositions. The desired quality in the mix is usually determined by quality control parameters, which are generally defined by ratios between different functions of the elemental components present in the mix. According to the present invention, the system controller periodically calculates the flow rate values to be applied to the different sources, then provides these values as set points for the local flow rates controllers for the sources. Each cycle of measurement and calculation is a "control instant." The time between control instants is a "control period." The set points for each control instant are determined are follows: (i) the on-line analyzer provides real time measurement of the composition of the mix; (ii) the system controller receives the measurement values and computes the mix quality control parameters; (iii) a plurality of advanced control loops are defined with one advanced control loop corresponding to each quality control parameter. For each advanced control loop, the variable to be controlled is the corresponding quality control parameter and the control signals are "advanced control signals" defined by ratios between functions of the flow rates of the different sources; (iv) an advanced control tool controls the advanced control loops and generates the values for the advanced control signals, (v) the advanced control signals are used along with the desired flow rate of mix to calculate the target flow rates for the different sources; and (vi) the new flow rate values are provided as set pointsto the flow rate controllers.

According to the present invention, direct closed loop control of the quality control parameters of the mix is provided by an advanced control tool which produces advanced control signals defined as ratios between functions of the flow rates to be applied to the different sources. Use of the relationship between the advanced control signals and the quality control parameters avoids the complexities and non-lineariies that arise in the relationship between the flow rates and the quality control parameters. Thus, in the context defined of the advanced control strategy of the present invention, the advanced control tool can perform a very precise control of the quality control parameters.

BRIEF DESCRIPTION OF THE DRAWINGS

The embodiment(s) of the invention are described in the Detailed Description with reference to the accompanying figures, wherein:

FIG. 1 is a diagram of the system according to the present invention for implementation of an advanced control strategy;

FIG. 2 is a block diagram showing the operations performed by the system of the present invention shown in FIG.1 ;

FIG. 3 is a diagram of an exemplary embodiment of the inventive system for use in a raw mix process for the production of the meal to feed a cement kiln;

FIG. 4 is a block diagram showing the operations perforaied by the embodiment of the system shown in FIG. 3 when a first set of quality control parameters is considered;

FIG. 5 is a block diagram showing operations performed by the embodiment of the system shown in FIG. 3 when a second set of quality control parameters is considered; and

FIGS. 6a-d are plots of the results obtained during 400 minutes of application of the process shown in FIG. 4 to a realistic simulation of the raw mix process, where FIG. 6a shows evolution of the quality control parameters considered and their corresponding set points; FIG. 6b shows evolution of flow rates computed according to the present invention; FIG. 6c shows evolution of the concentration of the main components in the different sources; and FIG. 6d shows evolution of the raw mix composition. DETAD ED DESCRIPTION OF AN EXEMPLARY EMBODIMENT

An exemplary embodiment of the system for raw mix optimization is shown in FIG. 1. It should be noted that the invention is not intended to be limited by the following detailed description of an exemplary embodiment for use in optimization of cement raw mix, and that adaptation of the inventive system and method to other blending processes will be readily apparent to those of skill in the art using the disclosure provided herein. As illustrated, the system for raw mix optimization comprises a plurality of hoppers 11 or other source containers with a means for feeding the contents of the container, a bulk material ingredient of the raw mix, onto a feeder conveyor 12. A flow rate controller (not shown) is provided at or in communication with the outlet of each hopper 11 to control the rate of flow of material out of the hopper and onto the feeder conveyor 12. Feeder conveyor 12 deposits the flow from each hopper onto the main conveyor 14 which carries the mixed material 15. A weigh scale 13 is associated with each feeder conveyor 12 to provide a measurement of mass per unit length of material on the conveyor 12. The measured mass per unit length is combined with the feeder conveyor speed to provide a value for the flow from each hopper 11 which is provided to the system controller for use in determining the advanced control values. The use of mass per unit length in the embodiment described herein is merely one way of calculating flow rate. Flow rate can also be determined using other mass or density measurements, as is known to those of skill in the art.

Main conveyor 14 carries the mixed material 15 into an elemental analyzer 16, which in the preferred embodiment is a PGNAA system such as that disclosed in U.S. Patents No. 5,396,071 (reissued as Re. 36,943) and No. 5,732,115, all assigned to the assignee of the present invention, or a similar on-line system that is preferably capable of providing real time results to permit mix corrections to be made quickly. The description of the system including a PGNAA system is intended to be illustrative, and the invention is not intended to be limited to analysis by prompt gamma tieutron activation. Generally, any on-line bulk material analyzer that is capable of measuring the composition of the bulk material may be used. Such elemental analysis methods may include, but are not limited to, x-ray techniques such as x-ray fluorescence and x-ray diffraction, and other neutron-based methods such as Thermal Neutron Activation/ Analysis and Fast Neutron Activation/Analysis.

It should be noted that main conveyor 14 does not terminate at analyzer 16, but continues beyond the analyzer, typically transporting the raw mix to a raw mill, a storage location, or in some cases, a kiln. Elemental analyzer 16 generates an output indicative of the elemental composition of the mixed material 15, which is input to the system computer/controller 17 for spectral analysis. For application to cement raw mix, the elements that can be analyzed may include calcium, silicon, iron, aluminum, magnesium, manganese, potassium, sodium, sulfur, and chlorine. System controller 17, which includes a computer such as an IBM-compatible PC, provides a number of functions, including spectral analysis computations and user interface functions, as well as processing the data obtained from the elemental analysis to determine the appropriate flow rates for the different bulk material sources. Programmable logic controller (PLC) 19 receives input from weigh scales 13 for communication to the system controller 17 and computes the appropriate motor drive signals from the flow rate commands generated by system controller 17. PLC 19 then provides the motor drive signals for flow rate adjustments from system controller 17 to motor control center (MCC) 18, which, in turn, transmits the flow rate control signals to the flow rate actuators. Where flow rates are computed from a feeder conveyor tachometer and the mass per unit length provided by the weigh scale 13, PLC 19 may either calculate the flow rates or provide the data to the system controller 17 to perform the calculations.

To perform the method of the present invention using the system of FIG. 1, the system is provided with n different sources (each in its own hopper 11), each of which will contribute to the produced mix with a flow rate that will be determined at each control instant k, where k is an integer which measures the time in control periods. Each source number i, Si (i = l,n), will contain different concentrations, Cij(k) (j = l,m), of m different components. The actual concentration of the different components in the measured mix, CA_j(k) (j = l,m), will directly depend on their actual concentration in the different sources, the actual flow rate of each source and the actual flow rate of the mix, according to the following general equation:

C (k) = C_s(k) x A FR(k), (1) where: C_AQ ) is a m-dimension vector, whose components are the actual concentrations CA_j(k) of the different components in the measured mix;

Cs(k) is a (m x n)-dimension matrix whose components are the concentrations of components Ci^l -D in the different sources;

Dj is an integer that represents in control periods the transport time delay from source i to the analyzer plus the measurement time delay of 1; and ANFR(k) is a n dimension normalized vector, whose components, ANFRι(k), are the n normalized flow rates, which determine the actual contribution of each one of the sources to CA(X), i.e.:

ANFR(k) = AFR,(k-D / Σ AFR,(k-D,), (2) ι = 1

where AFR_ι(k-D_ι) (i = l,n) are the actual flow rates from the different sources at instant k- D,.

In the raw mix process there is usually a qualitative, but not a precise quantitative, knowledge of the composition of the sources. In addition, this composition, i.e., the components of matrix Cs(k), may vary with time. Further, the actual flow rates AFR,(k) from the different sources may differ from the flow rate set points applied by the control system (FR^k)) due to the presence of process noises and/or inaccuracies in the performance of the local flow rate controllers. Finally, the actual mix composition vector C_A( ) may differ from the measured composition vector C_M(k) due to the presence of measurement noise, which is inherent in the analyzer measurement.

As previously described, the quality of the mix is generally determined by the value of "quality control parameters", also known as "quality control modules", of the mix composition. These quality control parameters are usually ratios in the following form:

QP,(k) = FQn,(Cι(k),C2(k), ...,C_m(k))

FQdjCCι(k),C₂(k), ...,C_m(k)) (i=l,QPN), (3) where FQn,( ) and FQd]( ) are functions of C k) j=l,m), which are the components of the measured composition vector CM(X), i.e., the percentage of the measured components in the mix. In some cases, FQ4( ) can be equal to 1. QPN is the maximum number of quality control parameters to be used.

The objective of an optimal control system for a raw mix process is to calculate the flow rates of the different sources at each control instant k in such a way that the quality control parameters approach their desired values within a minimum amount of time and remain as close as possible to their desired values throughout the process.

According to the present invention, the advanced control strategy defines one advanced control loop for each quality control parameter. The process variabb to be controlled in each advanced control loop is the corresponding quality control parameter. The corresponding advanced control signal is determined as the ratio between functions of the flow rates from the different sources. Because the different source feeders, e.g., feeder conveyors 12, are typically positioned serially along the main conveyor 14 so that they are each at different physical distances from the measurement point, there may be a different transport time delay D, from each source to the analyzer 16. In order to facilitate good performance of said advanced control loops, it may be convenient to establish a well defined time delay between each advanced control signal and the corresponding quality control parameter. This requires the related flow rate sources to have the same time delay in control periods with respect to the corresponding quality control parameter measurements. This result may be obtained by adding certain number of time delays to some of the flow rate sources. The certain number of time delays may be computed by making the following initial calculations:

First, the maximum delay MD, which is equal to the maximum (i⁼l_»n) delay, is calculated:

MD = max Di (4) l≤ i ≤ n

Then, an added delay vector AD, whose components, AJD[ (i⁼l _>n), is determined:

AD = MD- D, (i=l,n). (5)

The added delay vector components are used to apply an additional time delay to each corresponding source so that the delays are characterized in terms of the maximum delay. For systems with variable belt speed, changes in belt speed should be taken into account by recalculating maximum delay MD and added delay vector AD.

FIG. 2 illustrates the operations that the system of the present invention performs using the computer included in system controller 17 to implement the advanced control strategy at every control instant k. These operations are described as follows:

(a) in step 20, measurement of the mix composition is performed using on-line analyzer 16. The measurement is filtered in system controller 17 to obtain the measured composition vector CM(X);

(b) in step 21, system controller 17 computes the quality control parameters to be controlled Q^ (i=l,QPN) using Equation 3, in which the functions FQn( ) and FQd^ ) are determined for each particular application. The computed quality control parameters may be filtered if desired;

(c) in steps 22a - 22n, execution of the quality control parameters advanced control loops, QPACLOOP, (i=l, QPN) is performed using the advanced control tool. There will generally be an advanced control loop for each quality control parameter to be controlled. Each advanced control loop receives as a process variable the measurement of the corresponding quality control parameter, then uses the advanced control tool to compute the corresponding advanced control signal. Execution of the QPACLOOP,

(i=l,QPN) produces a set of advanced control signals, QPR (i=l,QPN), designated as steps 23a - 23n, which are applied to the process after being limit checked in step 24.

The advanced control tool executes each advanced control loop at every control instant k. The data used in execution of the loop include (1) the measured value of the quality control parameter corresponding to that quality control loop, (2) its set point, (3) the actual advanced control signal applied in the previous control instant k-1, and (4) the computed maximum delay (MD.) When FQd, is equal to 1, the advanced control tool may be provided with additional data as explained below;

(d) The set of the previously-computed advanced control signals will impose the following QPN conditions to the n flow rates for the n different sources:

QPR,(k) = Fn,(F₁(k),F₂(k),...,F,₁(k) )

Fd₁(Fι(k),F₂(k),...JF„(k)) (i=l,QPN), (6)

where Fnj( ) and Fd_ι( ) are functions of the advanced flow rates, F,(k) (i=l,n), to be applied to the different sources after a delay by a number of control periods equal to their corresponding added delay vector component ADi. Functions Fn,( ) and Fd,( ) are determined for each particular application. In cases where FQd;( ) is equal to 1, the corresponding Fd^) may also be equal to 1. If so, the advanced control tool can be provided with additional data from the actual values of the arguments of Fn,( ) applied to the raw mix process in order to execute the corresponding advanced control loop. (Note that additional data can be used even where FQd,( ) is not equal to 1, however, no significant advantage is gained.)

An additional condition on the advanced flow rates of the different sources is derived from the fact that the sum of the advanced flow rates must be equal to the desired flow rate of produced mix, as computed in step 26, i.e.: n

∑F(k)=MFRsp(k+MD), (7) ι = 1

where MFR_Sp(k+MD) is the set point for the flow rate of the produced mix at control instant (k + MD). From the (QPN + 1) conditions of Equations 6 and 7, the advanced flow rates F,(k) for each of the n sources are computed and limit checked. If (QPN + 1) is equal to n, computation of the flow rates will be unique. If n is less than (QPN + 1), which may occur for situations when a feed failure happens, it will not be possible to drive all the quality control parameters to their set points. In this case, a priority may be established to determine which QC parameters will remain under control (step 26). Also, if a feed failure happens, the control objectives may be changed and a different total flow rate of the mix may be used until the source failure is corrected.

In many cases, the number of sources will be greater than the number of quality control parameters +1, i.e., n > (QPN + 1). In such cases, a practical way of providing a unique solution is to identify (QPN + 1) sources as "preferential use" (PU) sources, leaving the remaining [n-(QPN+l)] sources as corrective sources that will only be used if needed. These corrective sources may be used to meet other criteria such as minimizing cost or reducing the need for materials that are difficult to grind or otherwise process for use.

(e) Using the previously-computed advanced flow rates, F(k), the actual flow rates to be sent at control instant k as set points to the local flow rate controllers, FR(k), are computed by:

FR_!(k)= Fι(k - ADi) (i=l,n). (8)

Thus, the actual flow rates to be applied as set points to the flow rate local controllers in step 27 will be equal to the advanced flow rates that were computed AD control periods ago.

In many cases, a convenient and practical procedure for applying the above- described general method is to select the preferential use (PU) sources such that each PU source becomes the primary supplier of a particular component of the components to be considered in the quality control parameters of the mix. Proper selection of the QPN conditions of equation (6), imposed by the advanced control signals of the advanced control loops, may allow development of a simple and efficient logic for diagnosis of deficiencies that can occur in some of the sources without the need for direct measurement or estimation of the individual source compositions.

The advanced control tool produces precise advanced control signals that will drive the quality control parameters to their set points when it is physically possible. If an advanced control signal remains at one of its absolute limits for an extended period of time, i.e., through many control instants, yet the corresponding quality control parameter still does not reach or tend to approach its set point, then the control objective is not feasible. An analysis of the advanced control signal definition as a ratio between functions of the sources flow rates provides a fault diagnosis in step 25, identifying composition deficiencies in the sources.

An exemplary application of the above-described general method for advanced system control is the raw mix process used in the cement production. A typical raw mix process of this kind is represented in FIG. 3 and is described as follows.

The primary components of interest in the raw mix to determine its quality are CaO, SiO₂, A1₂0₃ and Fe₂O_r The concentrations of these oxides in the mix are measured at a frequency of once per minute, thus defining the control period. Thus, flow rates for the different sources can be adjusted as frequently as every minute.

The quality control parameters in the raw mix may be defined in different ways, but in many cases are the Lime Saturation Factor (LSF), the Silica Module (SM) and the Iron Module (IM), which are usually defined by the following equations:

LSF = 100 CaO

(2.8 SiO, + 1.18 Al O + 0.65 Fe O, ) (9)

SM= SiOo

(Al₂0 +Fe^~0₃) (10)

An alternative choice for the quality control parameters may be the SM previously defined in combination with the following Bogue formulas:

C3S = 4.071 CaO - (7.602 SiO₂ + 4.479 A1₂0₃ + 2.859 Fe₂O₃ + 2.852 S0₃) (12)

C3A = 2.65 A1₂0₃ - 1.692 Fe₂O₃ (13) As shown in FIG. 3, six different sources are considered in this example. Sources 31, 32, 33 and 34 are selected as "preferential use" (PU) sources and as main suppliers in the mix of Ca, Si, Al and Fe oxides, respectively, while sources 35 and 36 are considered "corrective sources", which are used when an insufficient composition in the PU sources is diagnosed. The sources are described as follows:

Source 31 is limestone and has a high CaO content. This will be the main source used to supply the CaO required to control the LSF module.

Source 32 has a high SiO₂ content and low A1₂0₃ content. This will be the main source used to supply the SiO₂ required to control the LSF and SM modules.

Source 33 is marl and has a high SiO₂ content, high CaO content, and relatively high Al₂O₃ content. Due to its relatively high alumina content, this will be the main source used to supply ALO₃ and, consequently, all the QC modules will be affected by it. It may happen that the content of aluminum in this source is not high enough to attain the control objectives.

Source 34 has high Fe₂O₃ content. This is the main source used for the supply of Fe₂O₃ required to control the IM module.

Source 35 has high Si0₂ content and relatively high Al₂O₃ content. Source 35 is designated a "corrective source" and, therefore, will not be used until a deficiency of aluminum content in source 33 is detected.

Source 36 is sand and has a very high SiO₂ content. Source 36 is also designated a corrective source and may be used to correct a deficiency in the silica contents source 32.

As described in the general case, in the raw mix process of this illustrative example there are measurement noises, process noises and/or local flow rate controller inaccuracies and delays D; (i=l,4). The precise composition of the sources is unknown and time-varying.

The implementation of raw mix optimization according to the present invention in this illustrative example requires the initial computation of the maximum delay MD and the added delay vector AD, which is determined from the delays D; (i=l,4) corresponding to the selected preferential use sources, as previously considered in Equations 4 and 5. As defined in the general case, the delays D; are equal to the corresponding transport time delay from the source to the analyzer, measured in control periods, plus 1. For example, a source with a transport time delay of 135 sec. (2 min., 15 sec), has a transport time delay of three control periods and a D equal to 4. The implementation of operations (a) - (e) (see description for FIG. 2) for this illustrative example, assuming that the quality control parameters chosen are LSF, SM and IM, is illustrated in the schematic diagram of FIG. 4 and established at each control instant k in the following (aj) to (ei) steps, designated in the figure as steps 41-45:

Step 41: (aj) Measurement of the concentration of CaO, SiO₂, A1₂0₃ and Fe₂O₃ in the mix by the on-line analyzer and filtering of these values to obtain the measured composition vector CMO :

C^k)^τ = [CaO(K) SiO₂(k) Al₂0₃(k)Fe₂O₃(k)] (14)

Step 42: (bi) Computation of the quality control parameters, LSF, SM and IM, by means of:

LSF(k) = 100 CaO(k)

(2.8 SiO₂(k) + 1.18 A10₃(k) + 0.65 Fe₂O₃(k)) (15)

SM(k) = SiOJk)

(Al₂0₃(k) + Fe₂O₃(k)) (16)

IM(k) = __ _

Fe₂O ) (17)

These values may also be filtered if desired.

Step 43: (ci) Execution of the corresponding quality control parameters advanced control loops LSFACLOOP, SMACLOOP and IMACLOOP, by means of an advanced control tool ACT. The application of an advanced control tool in the present context can be carried out in a number of different ways which are known to those of skill in the field of computer engineering. In the instant example, three advanced control loops are initially created within the advanced control tool. Then, at each control instant k, the advanced control tool is invoked to execute each of the loops after receiving the appropriate data. This execution may be described by the following equations:

LSFR(k) = ACT(LSFACLOOP, "LSF(k)", "LSF_SF(k)", "ALSFR(k-l)") (18) SMR(k)=ACT(SMACLOOP, "SM(k)", "SM_SP(k)", "ASMR(k- 1 )") (19)

IMR(k) = ACT(IMACLOOP, "LM(k)", "lM_SP(k)", "AIMR(k-l)") (20)

In each of the above equations the advanced control tool (ACT) receives four arguments within the parenthesis. The first one is the name of the loop to be executed; the second and third are the corresponding quality control parameter and its set point computed at control instant k. The fourth is the actual advanced control signal applied to the loop in the previous control instant lc-1. In return, the advanced control tool delivers the advanced control signal at control instant k for each loop, which appears on the left- hand side of the above equations. As previously explained, these advanced control signals represent a ratio between two functions related to the flow rates to be applied to the sources. Many different types of advanced control tools can be used in the present invention, including neural net adaptive controllers, model reference adaptive controllers, self-tuning regulators, and adaptive predictive control systems. Adaptive predictive control systems are well known in the field of control systems. See, e.g., J.M. Martin Sanchez and J. Rodellar, Adaptive Predictive Control: From the Concepts to Plant Optimization, Prentice-Hall International Series in Systems and Control Engineering, 1996. An example of such a system is described in Patent No. 4,197,576 of the present inventor.

Step 44: (di) The previously-computed advanced control signals will impose the following three conditions to the flow rates to be applied to the sources: LSFR(k) = Fι(k)

(F₂(k) + F₃(k)) (21)

IMR(k) = F_β k)

F₄(k) (23)

The above conditions determine the functions Fn,( ) and Fd,( ) considered in step (d) above. F,(k) (i=l,4) are the advanced flow rates to be applied to the different PU sources after they are delayed by a number of control periods equal to their corresponding added delay vector component ADi. The additional condition expressed in Equation 7 in this case becomes:

Fι(k) + E;(k) + F₃(k) + F₄(k) = MFRs_P(k + MD) (24)

Where MFRs_P(k + MD) is the set point of the mix to be produced at control instant k + MD, which is the time in which the four advanced flow rates F k), F₂(k), F₃(k) and F₄(k) will be mixed and their mixed composition measured. From Equations 21 to 24, the advanced flow rates Fι(k), F₂(k), F (k) and F₄(k) will be computed and limit checked. Step 45: (ei) From the previously-computed advanced flow rates, Fι(k), F₂(k), F₃(k) and F₄(k), the actual flow rates to be sent at control instant k as set points to the local flow rates controllers, FRι(k), FR₂(k), FR^k) and FR₄(k), will be computed as previously defined in Equation 8. Thus the actual flow rates will be equal to the advanced flow rates computed ADi control periods ago, respectively.

In the preceding example, the choice of the preferential use sources and the definition of the advanced control loops illustrate the manner in which deficiencies in the source compositions that may prevent the attainment of the control objectives can be diagnosed. It may be preferable to first ensure that the control objectives in terms of LSFsp, SM_SP and IMsp are feasible. They are feasible if the following conditions are true:

1) LSF_sl(k) > LSF_Sp(k) and LSF_(S2+s₃₎(k) < LSF_SP(k) V k;

2) SM_S2(k) > SM_SP(k) and SM_S3(k) < SM_SP(k) V k, and

3) IMs₃(k) > IMsp(k) and IM_s (lc) < IM_SP(k) V k , where LSFsι(k) and LSF(_S2+s₃₎(k) are the LSF value of source 31 and the highest LSF value of any possible combination between sources 32 and 33, respectively, at control instant k. Since source 31 is the main supplier of CaO, LSFs{k) is the maximum possible value of any possible produced mix and, therefore, it must be higher than LSFsp(k). On the other hand, since sources 32 and 33 are respectively the main suppliers of SiO₂ and Al₂O₃, respectively, any of their combinations should result in a low LSF value of the possible produced mixes. Therefore, LSFs_P(k) should be lower than LSFsι(k) and higher than LSF(s₂₊s₃)(k) in order to ensure that the LSF control objective is feasible.

Similar reasoning to that previously applied for condition 1) may be extended to justify conditions 2) and 3), taking into account that at control instant k: (i) SMs₂(k) and SMs₃(k) are respectively the SM values of sources 32 and 33; (ii) IMs k) and IMs₄(k) are respectively the IM values of sources 33 and 34, and (iii) source 32, source 33 and source 34 are respectively the main suppliers of SiO₂, Al₂O₃ and Fe₂O

If one or more of the advanced control signals remains at one of its limits (either high or low) for a certain period of time and the set point of the advanced control loop cannot be attained by the corresponding quality control parameter, this control objective is not feasible. A simple exercise can be performed to determine the cause for the unfeasibility of the control objective after which an output can be generated with the corresponding diagnosis as shown in the following examples: Example A.- LSFR(k), defined by Equation 21, has been sitting at its highest limit for an extended time, which means that the LSF of the mix is primarily determined by the LSF of source 31. However, in spite of this condition, LSF(k) remains lower than LSFsp(k) and does not show any tendency to converge towards it. The diagnosis is that the LSF of source 31 is insufficient, i.e., there is not enough calcium oxide in source 31. The remedial options in this case would be to re-evaluate to determine whether source 31 was correctly selected and, if not, select another preferred use source with a higher concentration of CaO to bring up the LSF, or to use a corrective source to boost the CaO concentration in the mix.

Example B.- SMR(k), defined by Equation 22, has been hitting its lower limit (usually zero) for an extended time, which means that the SM of the mix is primarily determined by the SM of source 33. In spite of this condition, SM(k) remains higher than SMs k) and does not show any tendency to converge towards it. The diagnosis is that the SM of source 33 is too high, i.e., there is not enough aluminum oxide in source 33. A recommended remedial action in this case would be to use a corrective source with a higher aluminum oxide concentration, if available.

Example C- IMR(k), defined by Equation 23, has been hitting its upper limit for an extended period, which means that the IM of the mix is dominated by the IM of source 33. In spite of this condition, IM(k) remains lower than IMs_P(k) and does not show any tendency to converge towards it. The diagnosis is that the IM of source 33 is not high enough, i.e., the ratio between aluminum oxide and iron oxide in source 33 is too low. A recommended remedial action in this case could be to use a corrective source with a higher concentration in aluminum oxide, if available.

FIG. 5 illustrates the advanced control strategy (steps (a) to (e)) in a case where the chosen quality control parameters are C3S, SM and C3A. The MD is calculated along with the corresponding added delay vector. For each control instant k, steps (a:) to (e₂) steps are designated in the figure as steps 51-55:

Step 51: (a₂) This step is equivalent to step (ai) (Step 41 in FIG. 4) above.

Step 52: (b₂) Computation of the quality control parameters, C3S, SM and A3A, is performed using the following equations:

C3S(k) = 4.071 CaO(kH7.602SiO₂(k)+4.479Al₂O₃(k)+2.859Fe₂O₃(k)+2.852SO₃(k))

(25) SM(k) = SiO₂(k) / (Al₂0₃(k) + Fe₂O₃(k)) (26)

C3 A(k) = 2.65 Al₂0₃(k) - 1.692 Fe₂0₃(k) (27) As in the previous examples, these values may also be filtered if desired.

Step 53: (c₂) Execution of the corresponding quality control parameters advanced control loops C3SACLOOP, SMACLOOP and C3AACLOOP, is performed using the advanced control tool ("ACT"). In this case, the quality control parameters C3S and C3 A are instances of Equation 3 where the corresponding FQd;( ) are equal to 1. Thus, the corresponding Fd,( ) of Equation 6 will also be equal to 1, and execution of the advanced control loops will be performed in a manner similar to that of step (ci) (Step 43 in FIG. 4) above. However, in this case, the advanced control tool may receive additional data in the execution of loops C3SACLOOP and C3AACLOOP as described by the following equations: C3SR(k)=ACT(C3SACLOOP,"C3S(k)","C3S_SP(k)","AC3SR(k-l)","F2(k-l)","F3(k-l)")

(28) C3AR(k)=ACT(C3AACLOOP,"C3A(k)","C3A_Sp(k)","AC3AR(k-l)","F3(k-l)") (29)

In each of the preceding equations, the ACT first receives the four arguments already described in step (cj.) (Step 43 in FIG. 4) for Equations 18 to 20. Then, for the execution of advanced control loop C3SACLOOP, the ACT receives additional data comprising the actual advanced flow rates of sources 32 and 33 applied to the raw mix process at control instant (k-1), as shown in Equation 28. In a similar manner, for the execution of the advanced control loop C3AACLOOP, the ACT may receive additional data comprising the actual advanced flow rate of source 33 at control instant (k-1), as shown Equation 29. In response, the ACT delivers the advanced control signal at control instant k for each loop, which appears on the left hand side of the above Equations 19, 28 and 29.

Step 54: (d₂) The previously-computed advanced control signals will impose the following three conditions on the flow rates to be applied to the sources:

C3SR(k) = F_!(k) (30)

SMR(k) = F k) / F₃(k) (31)

C3A(k) = F₄(k) (32)

The above conditions determine the functions Fn;( ) and Fd,( ) considered in step (d) above. As in step (di) (Step 44 of FIG. 4), an additional condition expressed in Equation 24 must be satisfied. From Equations 30 - 32 and 24, the advanced flow rates Fι_,(k), F₂(k), F₃(k) and F₄(k) will be computed and limit checked.

Step 55: (e₂) This step is equivalent to step (ei) (Step 45 in FIG. 4) described above. In this case, the control objectives in terms of C3SSP, SMSP and C3A, are feasible if the following conditions are verified:

1) C3Ssι ≥ C3S_Sp and C3S_(S2+S3) ≤ C3S_SP;

2) SM_S2 > SM_Sp and SM_S3 < SM_SP) and

3) C3A_S3 > C3A_SP and C3A_S ≤ C3A_SP.

If one or more of the advanced control signals reaches its limits and the set point is not attained by the corresponding QC module, this control objective is not feasible. A similar logic to that previously considered in Examples A-C can be used to determine the reason for the inability to attain the control objective, issue a corresponding diagnosis, and suggest remedial action.

EXPERIMENTAL EXAMPLE

The system and method of the present invention have been used to control a simulation of the typical raw mix process. The basic software structure for this implementation used MATLAB^® (from Mathworks, Inc., Natick, Massachusetts), and included a simulation of the typical mix process shown in FIG. 3 and an implementation of the system controller, according to the schematic diagram of operations shown in FIG. 4 and described in points (ai) to (ej). Each execution of the MATLAB^® program represented a control instant. The particular details of the process simulation, the inventive method and the experimental results are described as follows.

The simulation of the mix process is built based on Equations 1 and 2, where the components of matrix Cs(k) are the concentrations of the oxides in the different sources. Some initial values for the concentrations of the oxides in each source are first considered at k = 0, however, these values change with k as "random walks" generated by integration of a MATLAB^® function called "randn." This function generates a gaussian noise with zero mean and a selectable standard deviation. The standard deviation for each of these "randn" functions is selected to be proportional to the corresponding initial composition values. The actual flow rates AFR,(k) from the different sources differ from the flow rate set points FR(k) applied by the inventive system due to a gaussian noise added to each one of them in the process simulation. The simulated mix process is normalized in the sense that the flow rate of produced mix, MFRsp(k+MD), is assumed to be equal to 1. Therefore, the sum of all the F_ι(k), considered in Equation 7, is made equal to 1. Also the mix process simulation adds a gaussian noise to each of the components of the actual mix composition vector C_\(k) in order to obtain the simulated measured composition vector C_M(k). Tables 1 and 2 present the specific details of the mix process simulation. Table 1 provides 1) initial source compositions, 2) initial flow rates, 3) delays Di of the sources, and 4) standard deviation of the gaussian noises on the different flow rates. Table 2 provides 1) standard deviations of the gaussian noises that generate the random walk changes in the oxide contents of the different sources, 2) standard deviation of the gaussian noises on the raw mix composition measurements for 1200mm band considered in this raw mix process simulation, and 3) discontinuous changes in the composition of the sources. The Table indicates oxide, source and instant where the change takes place.

The MATLAB^® implementation of the inventive raw mix optimization system obtained at each control instant k the measured values of the oxide concentration in the mix from the process simulation. Using the measured values, according to steps ( ) and (bi), the quality control parameters values were computed. Next, the MATLAB^® program communicated with an Active X control in order to execute the three advanced control loops, as considered in step (ci). Once the Active X controller communicated the advanced control signals to the MATLAB program, the flow rates to be applied were derived according to steps (di) and (ei). These flow rates were used as simulated raw mix process inputs in the next execution of the MATLAB program. For the test, the software included a routine for diagnosing deficient composition in source 33, such as that described in Example B above. Additionally, a routine was included to simulate an increase in the flow rate of corrective source 35 when the diagnosis routine was activated and to decrease it when the use of this corrective source was not longer necessary.

In the experimental results shown in FIGS. 6a-d, the raw mix optimization system starts automatic operation at instant k = 10. Table 3 shows initial values of the QC modules and changes in the modules set points. The new set point value and the instant when the change takes place are specified.

Table 1

Table 3

FIGS. 6a-d are four graphs showing the evolution of the following variables over a 400 minutes period of experimental application of the inventive system and method for raw mix optimization:

Beginning on the y-axis and proceeding top to bottom, FIG. 6a shows the evolution of the LSF module and its set point, the SM module and its set point, and the IM module and its set point with time. (Note that the values of the LSF module and its set point are divided by 10 in this figure to allow their representation with the same scale than that of the SM and the IM modules.) In each case, the set point for each module is designated by the dotted line. Additionally, the plot of FIG. 6a shows the activation of a diagnosis, designated D_AL, of a deficiency in the aluminum oxide content of source 33, which appears before instant 150 and disappears after instant 200.

In a similar manner, FIG. 6b shows, from top to bottom, the evolution of the flow rates for source 31, source 33, source 32 and source 34, with each line of the plot designated by the reference numeral for the corresponding source. In addition, the evolution of the flow rate for corrective source 35, which starts to increase from zero before instant 150 and goes back to zero before instant 350.

FIG. 6c shows the evolution of the compositions in the sources. Proceeding from top to bottom, compositions are plotted for SiO₂ in source 32, CaO in source 31, Fep₃ in source 34 and A1₂0₃ in sources 35 and 33.

Filially, FIG. 6d shows the mix composition measurements with time for calcium (Ca), silica (Si), aluminum (Al) and iron (Fe) oxides in order from top to bottom.

The experimental results plotted in FIGS. 6a-d illustrate the performance of the system and method for raw mix optimization when submitted to set point changes in the quality control parameters, stochastic and discontinuous changes in the sources' composition, flow rate noise acting on the flow rates and raw mix noise acting on the mix oxide composition measurements. Further, they illustrate the performance of the inventive system and method when a deficient composition in the PU sources is produced and a corrective source is used to achieve the quality control parameters set points. The following observations can be made regarding performance of the inventive system and method based on analysis of FIGS. 6a-d at sampling instants at which conditions change:

At instants 20, 40, 60 and 100, which are indicated in FIG. 6a, the inventive method shows satisfactory performance for responding to and achieving set point changes on LSF, SM, IM, and LSF again, respectively. While the new target set point is attained for each module for which the change is desired, Hie measurements for the other modules are unaffected by the change, remaining at their original set points. For example, at instant 30 (approximately), the mix has been changed to allow LSF to reach the new set point. Even with the change in the mix, SMand IM remain relatively steady at their original set points, at least up until instant 100. The last set point change in LSF at instant 100 is achieved in the presence of a discontinuous change in the aluminum oxide concentration of source 33.

At instant 100, the content of aluminum in source 33 is diminished from 10.87 to 6.87. (See FIG. 6d.) It may be observed from FIG. 6b that in attaining the control objectives, the flow rates of the sources are changed by increasing source 33 and decreasing sources 31 and 32. However, because the SM module continues to deviate from its set point, the system activates a diagnosis of deficient aluminum oxide concentration in source 33 and begins adding material from corrective source 35 at around instant 140. The use of this corrective source with high content of aluminum allows the SM module to return to its set point by mstant 200. The system controller provides an output of the diagnosis. At around instant 250, the SM set point increases from 2.3 to 2.6. At tfris point, less aluminum is necessary to maintain the control objectives and the system gradually decreases the contribution of material from the corrective source 35, until the flow rate is zero. As shown in FIG. 6b, at instant 325, the content of aluminum in source 33 is restored, returning from 6.87 to 10.87. Comparison of the plotted flow rates in FIG. 6b and the QC modules in FIG. 6a illustrates how the system and method decrease source 33 and increase sources 31 and 32 with little impact on the evolution of the QC modules.

Accordingly, the present system and method for raw mix optimization provide precise control of the quality control parameters in the presence of changes in the control objectives (set points changes), high levels of measurement and process noises and perturbations, such as random and discontinuous changes in the source compositions. Further, the system and method provide diagnoses of insufficient composition within the available (preferred) sources to attain the control objectives and, when necessary, draws upon corrective sources to compensate for deficient compositions in the preferential use sources. The system and method of the present invention address a long-standing need to overcome control difficulties for the optimization of cement and other raw mix production processes, providing satisfactory performance in the time-varying, noisy environment typical of the raw mix industrial process without requiring time-consuming tuning by the user.

Other embodiments, applications and modifications of the present invention may occur to those of ordinary skill in the art in view of these teachings. Therefore, the scope of the invention is to be limited only the appended claims which include all other such embodiments and modifications when viewed in conjunction with the above specification and accompanying drawings.

Claims

What is claimed is:

1. A system for optimizing composition of a mixed material comprising: a plurality of sources for feeding a raw material for creation of the mixed material, each source having a flow controller with an adjustable flow rate set point; a bulk material analyzer for measuring composition of the mixed material at each of a plurality of control periods; a system controller for processing composition measurements from the analyzer to calculate a plurality of quality control parameters for the mixed material, each quality control parameter having a target value, defining a quality control loop for each quality control parameter, and generating for each quality control loop an advanced control signal to make the corresponding quality control parameter to converge towards its target value; wherein at least one of the advanced control signals comprises a ratio between functions of the flow rate for at least two of the plurality of sources; the at least one advanced control signals being used to adjust the flow rate set points of the source flow controllers at each of said plurality of control periods.

2. The system of Claim 1 , wherein the system controller further calculates a time delay for delaying adjustment of the flow rate set point based upon a transport time between the source and the analyzer.

3. The system of Claim 1 , wherein each quality control parameter is determined according to the relationship:

QP_ι(k) =

where FQn_ι( ) and FQd,( ) are functions of C_j(k) (j=l,m), which are components of a measured composition vector CM( ) that correspond to the percentage of the measured components in the mixed material and QPN is the maximum number of quality control parameters.

4. The system of Claim 1 , wherein the plurality of sources comprises a combination of preferred use sources and corrective use sources, and the advanced control signals are generated for controlling the preferred use sources, and wherein system controller further diagnoses a deficiency in the preferred use sources based upon a failure to attain the target values.

5. The system of Claim 4, wherein the system controller, upon diagnosing a deficiency in the preferred use sources, provides a control signal for controlling the flow rate of at least one corrective source.

6. The system of Claim 1, wherein the mixed material is cement raw mix and the quality control parameters are lime saturation factor (LSF), silica module (SM) and iron module (IM).

7. The system of Claim 1, wherein the mixed material is cement raw mix and the quality control parameters are silica module (SM), C3S and C3A.

8. A method for optimizing composition of a mixed material comprising:

(a) providing a plurality of sources for feeding a raw material for creation of the mixed material, each source having a flow controller with an adjustable flow rate set point for adjusting flow rate;

(b) measuring composition of the mixed material at each of a plurality of control periods;

(c) calculating a plurality of quality control parameters from the measured composition;

(d) defining a quality control loop for each quality control parameter;

(e) generating for each quality control loop an advanced control signal adapted to make the corresponding quality control parameter converge toward its target value, wherein at least one of the plurality of advanced control signals comprises a ratio between functions of the flow rate for at least two of the plurality of sources; each advanced control signal being used to adjust the flow rate set points of the source flow controllers at each of the plurality of control periods.

9. The method of Claim 8, further comprising calculating a time delay for delaying the application of the flow rate set point based upon a transport time between the steps of feeding the raw material and measuring composition of the mixed material.

10. The method of Claim 8, wherein each quality control parameter is determined according to the relationship:

QPi(k) = FQn,(C,(k),C₂(k), ...,C_m(k)) FQdKCiGO Ck), ...,C_m(k)) (i=l,QPN), where FQn;( ) and FQd;( ) are functions of Cj(k) (j=l,m), which are components of a measured composition vector C_M( ) that correspond to the percentage of the measured components in the mixed material and QPN is the maximum number of quality control parameters.

11. The method of Claim 8, wherein the plurality of sources comprises a set of preferred use sources and a set of con-ective use sources, and the advanced control signals are generated for controlling the preferred use sources, and further comprising diagnosing a deficiency in the preferred use sources if the quality confrol parameters fail to attain the target values.

12. The system of Claim 11, further comprising providing a control signal for controlling the flow rate of at least one corrective source following the step of diagnosing a deficiency in the preferred use sources.

13. The method of Claim 8, wherein the mixed material is cement raw mix and the quality control parameters are lime saturation factor (LSF), silica module (SM) and iron module (IM).

14. The method of Claim 8, wherein the mixed material is cement raw mix and the quality control parameters are silica module (SM), C3S and C3A.