CN1361079A - Method and apparatus for producing cement - Google Patents

Abstract

The control method of milling system in cement mill includes: inputting the set points and limit values of controlled and managed variables via the figure supervising and control interface (SCI) to data collection software interface; associating the controlled variables to operated variables in MPC, providing process control intervention (PCI) software module; providing post-processor rules based on overrunning software module; comparing signals from MPC and/or PCI to generate operation to correction signal; and sending the operation or correction signal to data collection software interface and converting into commands for operating or correcting the parts of the grinding system.

Description

The method and apparatus that cement is made

The method and apparatus that the relevant cement of the present invention is made.In the current embodiment of the present invention, the relevant a kind of method and apparatus of making cement of the present invention especially adopts a kind of wet method or dry way process to make the method and apparatus of cement with a kind of rotary kiln.One special aspect, operation and control that the present invention is undertaken by the system that grinds about the order of the order process by a kind of process steps in the cement mill to grinding operation.

Conventional cement manufacturing processed comprises that lime material (for example Wingdale, marlstone, chalk and analogue thereof), clay material (for example clay, shale, slag, coal ash, (volcanic ash) sand and analogue thereof) and/or siliceous material with predetermined proportion mix, and form all these material mud or that do also is sent to and imbeds cement formation kiln.The grog that is produced is mixed and grinds and form usually the modern silicate cement of dry powder powder mutually with gypsum.Ratio of each batching, their fine grinding degree have determined the chemical ingredients of grog and last cement products in the mixture.Additive is used for becoming phase-splitting to mix with one or more special cement feature is provided.Control process is purely by being realized that by the determined manual method of different operating person's decision therefore most manufacturing processed is the suboptimization operation basically.

One key operation in the cement mill is the operation of the system of grinding.One in a cement mill grinds system, the especially broken system of a dry grinding composition one ball mill, vertical roller press (VRM) or roller press (RP) is arranged, be used to receive a parallel feeding that is grouped into by several one-tenth, each of these compositions is dumped on the grinding machine feed conveyor belt via feeding warehouse separately.Those can not then be sent in grinding machine feed trunnion or the discharging chapelet with pneumatic mode by the composition (for example dry coal ash) that a travelling belt is fed.Ball mill during cement grinds is air-sweeping type typically, and one exhausting air-flow that wherein adopts the suction effect of means under the gas blower to be produced passes through grinding machine with materials conveyance.Material in the grinding machine is unloaded to a bucket type lift and is transferred into a separator, a dynamic air blowing separator normally.In separator, material is divided into the roughage (coarse fodder) that is back to grinding machine and refaces and is sent to the fines that bin goes.Any fine dust in the grinding machine exhausting air-flow also is collected in a rod-curtain precipitator (ESP) or the deep bed filter simultaneously and the fines with separator mixes mutually.

Grinder system is an integral part of device for producing cement.

Quality and final cement product quality are vital to grind operation and is subjected to the periodic disturbances that rule produces and influences to grinding, and these periodicity interferences have influence on the output of factory and production process.Each parts of grinder system have the control loop of himself and are used to keep the monitoring capability and the modulability of a given set(ting)value.

The objective of the invention is for cement mill provides a kind of system that grinds, the performance that wherein grinds system is controlled such that the total feed amount of grinding machine becomes maximum, compensates any process skew, guarantees that simultaneously product (fines) satisfied the fineness index of range request.The fineness typical earth surface is shown a kind of particle size distribution to rough grinding machine, and to the Brian specific surface area of cement mill.Other process variable that should be kept within operation and/or safety margin is grinding machine accumulated amount, grinding machine power of motor, chapelet power, separator power and circulation load (the total feed amount is removed by the new amount of feed) or coarse fodder ton amount.

Up to now, these operations of the system that grinds in the cement mill are mainly controlled by the operator.The invention provides a kind of dry type that is used in the cement mill and grind system, this system is provided with a kind of multivariate model based on supervision and control.The purpose of this method is that the control system that grinds improves output to reach, quality product is remained on the required technical indicator, and the compensation process deviation also guarantees all devices operating parameters is remained within the predetermined safe limit.The functional of control is to be on the higher supervision level, and is performed as through a distribution control system (DCS) or programmable logic controller (PLC) network target is offered the control loop that the proportional-integral derivative controller (PID) of low one-level comes each self-acting control to be associated with the system of grinding.

Control to the system of grinding is to send into the virgin material tonnage of grinding machine, the revolution of separator and the air door that grinds on the different blowing lines in the system (typically be grinding machine and suck line and separator blowing line) by manipulation to realize.In monitoring method of the present invention, controller is according to the value that the measuring result and the performance index mentioned above of controlled variable are calculated all above-mentioned variablees simultaneously.By a line solver that is tied optimal problem is guaranteed optimum control, wherein requiring minimized objective function is sum of squares with respect to the performance index deviation, the variable handled then be formulated as hard constraint in the optimization as decision variable all operations and safety margin.

According to the present invention, provide a system that grinds that is used for the cement mill, the parts of this system have one group of predetermined control with the variable of handling, they are controlled by a kind of method that includes following step:

A supervisory control computer that one data acquisition software interface (SuC) is often arranged is provided;

A figure supervision and control interface (SCI) is provided;

The setting point and the constraint of the variable of Be Controlled and manipulation are input in (SuC) through (SCI), (SCI) offers measuring result one Model Predictive Control software module (MPC) conversely again, and this module has the multivariate state-space model of the operating structure that is used to each parts of the system of grinding;

In (MPC) controlled variable (CV) is associated with by manipulated variable (MV), this is based on by manipulated variable (MV) and is received as through data acquisition interface that the plant data of the feedback signal in (SuC) develops out.

A process control intervention (PCI) software module is provided, and this module uses the operating process data to come the estimating system product fineness as control calculated feedback data in (MPC);

A preprocessor rule based on override software module (RBO) is provided, and it and one group of rule based on operator's behavior synthesize overall and are applicable to from (MPC) and/or from (PCI) and receive control signal;

To receive signal from (MPC) and/or (PCI) makes comparisons and produces operation or gauged signal;

Signal this operation or gauged is sent to (SuC) is used to convert to instruction, these instructions are provided to operate or proofread and correct the whole of the system of grinding or any one parts simultaneously by distribution control system (DCS)/programmable logic controller (PLC).

According to a preferred embodiment of the invention, present method comprises by authenticating an enough big group model makes these models in (MPC) be applicable to that covering grinds the entire operation zone of system component, and each of this group model is applicable to a subclass of this operating area.

According to the present invention, MPC typically adopts a kind of weighted mean to all model outputs to be used for predictive control.Preferably, the weight of model is to come online dynamically calculating according to any one operating point constantly.

According to a preferred embodiment of the invention, this method comprises that the lab analysis data in the employing (PCI) refresh the step of fineness model parameter, and condition is to depart from the lab analysis value if the prediction of model is found.

Present method preferably includes to adopt currently provides (PCI) to determine disturbing (for example variation of feed fineness degree, humidity) whether affecting process for operator and the employed intervention rule of slip-stick artist.

Interfere information is especially made the behavior of the fast regular that is used to provide (SuC) with feed-forward mode by (MPC).Followingly narrate the present invention with reference to accompanying drawing, in the accompanying drawing:

Fig. 1 illustrates the different parts of this grinder system for according to the functional diagram that is used for the control method of cement mill system of the present invention.

Fig. 2 and Fig. 3 illustrate the typical case who is used for Portlandpozzolan cement grinder system shown in Figure 1 and control structure and the typical supervision and control process implemented according to the present invention.

Fig. 4～10 are the avatars of an example of the method according to this invention operation, illustrate that cement mill system throughput, quality product, input total feed material concern over time, the operation of accumulated amount per-cent and Brian estimation model.

With reference to these accompanying drawings, shown in the figure according to a kind of method and apparatus that is used to control a cement mill system of the present invention.

Fig. 1 in the accompanying drawing illustrates a typical cement grinding mill system 10.Cement grinding mill system 10 in a cement mill, the especially dry method system that grinds forms a ball mill, vertical roller press or roller press GM, be used for receiving a synthetic feeding material of several compositions from hopper 14,16 and 18, this feeding material can be grog, gypsum and volcanic ash (coal ash), and each in them tilts on the grinding machine infeed conveyors 18 through feeding warehouse 14,16 and 18 separately.

Those can not preferably can pneumatic mode be transmitted by the composition of charging (for example coal ash) easily through a travelling belt and enter in grinding machine feed trunnion or the discharging chapelet.Ball mill during cement grinds is air-sweeping type typically, and one exhausting air-flow that wherein adopts the suction effect of gas blower 12 outlet sections to produce passes through grinding machine with materials conveyance.

The material of grinding machine GM is unloaded to a bucket type lift 22 and is transferred into a separator, a dynamic air blowing separator 20 normally.In separator 20, material is divided into roughage (coarse fodder) CV1 that is back to grinding machine GM and refaces and is sent to the fines that bin 28 goes.Any fine dust among the grinding machine exhausting air-flow CV4 also is collected in a rod-curtain precipitator (ESP) or the deep bed filter 26 and with the fines of separator 20 simultaneously mixes mutually, and is executed with one and flow separator 24 and be drawn in bin 28 lines.

The objective of the invention is to control the performance index of the system that grinds 10 of cement mill, make the total feed amount MV4 that obtains grinding machine GM become maximum, guarantee that simultaneously product (fines) CV6 satisfied the fineness index of range request.The fineness typical earth surface is shown a kind of particle size distribution to rough grinding machine, and to the Brian specific surface area of cement mill.Other process variable that should be kept within operation and/or safety margin is grinding machine accumulated amount CV2, grinding machine power of motor CV3, chapelet power CV5, separator rotating speed MV3, and air door manipulated variable MV1 and MV2 and circulation load (the total feed amount is removed by the new amount of feed) or coarse fodder ton amount CV1.MV1～MV4 wherein is by manipulated variable, and CV1～CV6 is controlled/monitored variable.

Fig. 2 and Fig. 3 illustrate the design of controller.

This controller is based on a kind of mixture model forecast Control Algorithm, and is made up of three modules, hereinafter they is narrated.

Especially with reference to Fig. 2, this illustrates the supervision and control structure.Supervision and control is implemented by supervisory control computer SuC, and the operator imports this computer SuC by SCI with set(ting)value and constraint.Data acquisition interface among the SuC offers measurement program block PCI, RBO and MPC conversely again and receives control signals from those program blocks and is used for a correction/operational order DCS (the dcs)/PLC (programmable logic controller) that transmits scriptures is transferred to the different parts of masher 10, this DCS/PLC be provided with target and be used as controller SuC as shown in Figure 1 and each parts of masher system 10 between an interface.Model Predictive Control module (MPC)

The present invention is based on Model Predictive Control (MPC) technology, several versions of this technology have been awarded patent (people's such as Prett No. the 4th, 349,869, United States Patent (USP) (patent of equal value that does not have India) for example, nineteen eighty-two; The United States Patent (USP) of Lu the 5th, 572, No. 420 (patent of equal value that does not have India) 1996 years; No. the 5th, 659,667, people's such as Buecher United States Patent (USP) (patent of equal value that does not have India), 1997 years).MPC has also been carried out extensive studies and delivered several progresses (Garcia and Morshed to this in open source literature, 1986-" quadratic programming of dynamic matrix control is separated (QDMC) ", Garcia C.E. and MorshedA.M., the chemical industry communication, 46 phases, 73-87 page or leaf (1986); Garcia etc. ,-Garcia in 1989 C.E., Prett D.M. and Morari M., " Model Predictive Control: theory and practice-summary ", automatization, the 23rd the 3rd phase of volume, 335-348 page or leaf (1989); Ricker ,-Ricker in 1994 N.L., " the present art of Model Predictive Control ", the 4th the chemical process control meeting in the world of CPC IV-procceedings, Padre Island, Texas, 271-296 page or leaf (1991)).Adopt the multivariate state-space model that grinds system component according to MPC module of the present invention, controlled variable CV1～CV6 association is come manipulated variable MV1～MV4, development comes variable MV1～MV4 according to plant data.The adaptive entire operation zone that covers the system of grinding of model is to come bonded in addition by authenticating an enough big group model, and each model wherein is applicable to a subclass of this operating area.The MPC algorithm adopts a kind of weighted mean to all model output to be used for predictive control.The model weighted value is then dynamically calculated online according to the operating point in arbitrary moment.Rule-based override module (RBO)

RBO is the postprocessor module that is used for MPC.Its function be monitoring disturb or the alarm procedure condition and to not by process model institute rightly the situation of description take quick corrective action.RBO is a kind of on-line expert system, and closes as a whole based on the rule of common operator behavior.Process condition is intervened module (PCI)

Grinding at cement has several conclusive controlled variables in the system 10, they are not to be arrived by on-line measurement.The fineness CV6 of the finished product is most important of this class variable.The PCI module adopts process data to come the estimating system product fineness as being used for MPC control calculated feedback value.Also have a kind of measure to adopt the lab analysis data to refresh the fineness model parameter, condition is to depart from the lab analysis value if model prediction is found.In addition, this PCI module also comprises several current persons of being operated and slip-stick artist and employed being used to determines whether to disturb (for example variation of feed fineness degree, humidity) affecting the intervention rule of process.This interfere information is made the action that is used to provide fast regular by the MPC module in the mode that feedovers.

Fig. 3 illustrates according to hybrid control method of the present invention, is described below.Square CVPV representative be used for controlled variable (process variable (PV) value of CV1～CV6) and MVPV representative by manipulated variable (process variable (PV) value of MV1～MV4), and be read into by current setting point (SP) value of the MV variable of square MVSP representative.These values are at first monitored by exception condition igniter module ATC, the excessive deviates (this deviate will indicate these control setting points correctly do not used) of MV MV1～MV4 inspection between PV and SP all during this module are equally also checked excessive velocity of variation at conclusive CV CV1～CV6, if there is not unusual condition, so all MV, PV and SP are sent to a MV filter block MVF, and it filters the transient state characteristic that the MV value is considered the cement grinding mill system.

MV value that is filtered and CV PV value are sent to Brian estimator module BE, and it calculates the Brian estimated value that is used for finally being ground cement products.Exception condition triggering device ATC, comprised process condition intervention (PCI) module of Fig. 2 by manipulated variable filter block MVF and Brian estimator BE.All estimated values and PV value are used as process feedback square PFB then and offer Model Predictive Control (MPC) module.Process setting point that the MPC module is also provided the operator of factory and constraint are as importing.Normal MPC calculates to be carried out and produces optimum setting point (OSP).If detect an exception condition is arranged, then producing has an alarm AL and this MPC by bypass.In this example, all PV and SP value are supplied to rule-based super (RBO) module of sailing, and this module recovers fast from exception condition with a quick sampling frequency control cement mill system 10.Normal optimum setting point from the MPC module is also checked that by RBO constraint is by the infringement situation.The output of RBO is final MV SP value, and as if it is downloaded to that field device piece DCS/PLC influences masher system 10 all the control of the such parts of blowing line air door 12, masher lift amplifier 22 and separator rotating speed 20.Benefit

From implement this supervision and control technology obtainable benefit comprise: to the control of the robust multivariable of total system to the reliable and regular On-line Estimation of product fineness fast system stability obtain higher mean yield from system and keep the special energy expenditure that required fineness is reduced in the attrition process and significantly reduce the more stably operation novel feature of the mutability of output and product fineness all important process apparatus

New feature of the present invention described herein is to have adopted a kind of hybrid MPC technology, has wherein adopted a kind ofly rule-basedly super to sail that (RBO) control is monitored and aftertreatment is calculated the control of MPC module.This RBO forms has the dry method of using in the cement mill to grind very typical working rule the system.

Another new feature is that process condition is intervened module, has wherein adopted a kind of principal component model to come to estimate according to the observed value of other process variable in the system variation (particle size or Brian specific surface area) of product fineness.

New feature of the present invention can be summarized as following several then: have the expert systems and the online product fineness estimator of the hybrid MPC that intervenes module based on principle component analysis that are used to control the cement mill system

Core control method in the method according to this invention is Model Predictive Control (MPC).The controller model that is used among the MPC is a discrete state space representation to process:

x(k+1)＝φx(k)+Γu(k) (1)

X is a state variable vector in y (k)=Cx (k)+Du (k) (2) formula, and u is the vector of MV, and y is the vector of CV.K is a current time.

In control is calculated, based on the predicted output of k+1 constantly that is positioned at of current information

Be calculated as: $\hat{y}(k+1/k)=\mathrm{C\φ}\hat{x}(k/k-1)+[\mathrm{C\Γ}+D]u\left(k\right)$ $+[y\left(k\right)-\hat{y}(k/k-1)]---\left(3\right)$

In case predicted output is calculated, an optimization problem then is established to calculate and obtains required target desired control motion.Optimization step is formed one group of optimum control campaign finding out future, and they drop to minimum to the predicated error by the method for least squares gained in estimation range in the condition that satisfies the process constraint.This predicated error is calculated as follows: $e(k+j)=y^d(k+j)-\hat{y}(k+j/k)---\left(4\right)$ Y in the formula ^dBe required path (typically being the setting point of current time).

Minimized objective function is provided by following formula:

MinX (Δ u)=e ^TQe+ Δ u ^TThe constraint that R Δ u bears defined.To the process constraint of variable that handled and output can according to future controls movement constraint write out.R and Q are respectively the weight matrix of controls movement and predicated error.Weight matrix definite constituted turnover step important among the MPC, and their the turnover rule of selection has been discussed (Maurath etc., 1988 years by numerous investigators; Garcia etc., 1989).

For addressing the above problem, listed a quadratic programming (QP) subproblem, wherein output and be expressed as hard constraint to decision variable by manipulated variable constraint:

minq(Δu)＝1/2Δu ^TGΔu+g ^TΔu (6)

Constraint Cc ^TΔ u 〉=bc (7) G is the sea plug matrix (Hessian) of objective function (5), and g is a target function gradient.(7) represent the system of linear equality and inequality constraint.About fascicle is prescribed as follows.Retrained by manipulated variable

To acting on by the boundary constraint on the manipulated variable, Cc ^T=l _L, l in the formula _LBe a lower triangular matrix, wherein all nonzero elements are equal to+and 1.To a coboundary u _Max, lower boundary u _MinAnd by the currency u (k) of manipulated variable, the bc element is set as concerning lower bound and equals u _Min-u (k) is set as concerning the upper bound and equals u (k)+u _MaxConcerning constraint of velocity, Cc ^T=1,1 equals the value that maximum can allow controls movement for unit matrix and bc element are set as in the formula.Output constraint

Applying of output constraint then is that the predicted output of requirement is positioned within the required border on the subclass that whole estimation range or this estimation range one is referred to as to retrain window.J the element of bc is given: for a lower boundary y of output _Min, $\mathrm{bc}\left(j\right)=(y\min -y\left(k\right))-\underset{i=1}{\overset{N-1}{\mathrm{\Σ}}}(\mathrm{aj}+i-\mathrm{ai})\mathrm{\Δu}(k-i)---\left(8\right)$ A coboundary y for output _Max, $\mathrm{bc}\left(j\right)=(y\left(k\right)-y\max )+\underset{i=1}{\overset{N-1}{\mathrm{\Σ}}}(\mathrm{aj}+i-\mathrm{ai})\mathrm{\Δu}(k-i)---\left(9\right)$ a _iBe i element of the right step response vector of CV-MV, Cc ¹Respective element be the j row of dynamic matrix A, this matrix A be a following triangle Toeplifz matrix based on step-response coefficients, is by record CV the transient response that the step of MV changes generation to be obtained.

Following reference is narrated the present invention according to a real work example of the operation of Controlling System of the present invention, and this example is to implement in the factory of 1,000,000 tons of cement of family's annual output of India.Some result who obtains by this enforcement is described in hereinafter and comes out with diagrammatic representation in Fig. 4～10 of accompanying drawing.Three kinds of situations have been discussed.Situation 1 normal optimum control

The online data of Fig. 4,5 and 6 for being captured at monitoring controller SuC on period.Controller is to be brought into during by 18,15,10 minutes after the cold starting at grinding machine to work online, and this moment, process was in transient behaviour.In figure shown in Figure 4, the output after line a represents to start.This is faster required normal steady time under Artificial Control than system.Among Fig. 5, online fineness estimated value is represented as line b in the drawings.Under be limited to 330mz/kg, and handle when can see and be used to guarantee fineness is remained on higher output separator rotating speed line c.It is the desired combined feed total feed setting point of controller that Fig. 6 illustrates line d, and line e is (the belt travelling belt is measured gained) actual input rate.In 1 grade of PID control behavior, tangible deviation is arranged, and supervision and control also can compensate this error.Situation 2 exception conditions are recovered

Exemplify the function of RBO module herein.Among Fig. 7, line f illustrates controller is increasing system throughput, and have the slope of a sudden change this moment when grinding machine gathers in the line g (see figure 8) at 14:05, is that the unexpected increase by charging humidity produces.If the control behavior does not resist above-mentioned this situation, grinding machine will be blocked, thereby cause serious shutdown.Can see relatively with line i, controller is taked smoothly to proofread and correct behavior line h and is revolted this interference, and gathers to just often then taking optimum control when grinding machine.Under hand control, the operator typically can reduce inlet amount significantly, thereby produces production loss and quality product change.This in service, product fineness changes the boundary that does not exceed defined.Situation 3 Brian estimated performances

Figure 10 has illustrated the performance of the online Brian estimation model of PCI module of the present invention.The actual Brian specific surface area value that this illustrates product is by the lab analysis that the sample of gathering in 36 hours periods is carried out resulting (line j), also illustrates the quilt from PCI module gained in the identical sampling time is estimated Brian value (line k).Between these two lines be numerically again the degree of closeness aspect curvilinear trend set up the precision and the reliability of Brian estimation technique.

Claims (9)

1. the system that grinds in a cement mill, it has one group of predetermined variable controlled and that handled, it is characterized in that; The parts that grind system are controlled by a method, and this method comprises the steps:

One supervisory control computer that has a data acquisition software interface (SuC) is provided;

One figure supervision and control interface (SCI) is provided;

Be input in (SuC) through (SCI) with Be Controlled with by the setting point of the variable handled and binding occurrence, (SCI) offers measuring result one Model Predictive Control software module (MPC) conversely again, and this module has the multivariate state-space model of the operating structure that is used to each parts of the system of grinding;

In (MPC) controlled variable (CV) is associated with by manipulated variable (MV), this is based on by manipulated variable (MV) and is received as through data acquisition interface that the plant data of the feedback signal in (SuC) develops out;

A process control intervention (PCI) software module is provided, and this module uses the operating process data to come the estimating system product fineness as control calculated feedback data in (MPC);

A preprocessor rule based on override software module (RBO) is provided, and it and one group of rule based on operator's behavior synthesize overall and are applicable to from (MPC) and/or from (PCI) and receive control signal;

To make comparisons from the signal that (MPC) and/or (PCI) receives and produce operation or gauged signal; And

Signal this operation or gauged is sent to (SuC) is used to convert to instruction, these instructions are provided to operate or proofread and correct the whole of the system of grinding or any one parts simultaneously by distribution control system (DCS)/programmable logic controller (PLC).

2. one of control cement τ as claimed in claim 1 grinds the method for the parts of system, it is characterized in that: this method comprises by authenticating an enough big group model makes these models in (MPC) be applicable to that covering grinds the entire operation zone of system component, and each of this group model is applicable to a subclass of this operating area.

3. one of control as claimed in claim 1 or 2 cement mill grinds the method for the parts of system, it is characterized in that: MPC adopts a kind of weighted mean to all model outputs to be used for predictive control.

4. grind the method for the parts of system as one of one of above-mentioned claim described control cement mill, it is characterized in that: the weight of model is to come online dynamically calculating according to any one operating point constantly.

5. grind the method for the parts of system as one of one of above-mentioned claim described control cement mill, it is characterized in that: disturb or alarm condition come by (PCI) responsive, should (PCI) will (MPC) bypass warn (RBO) be used for to those not by the process model of (MPC) institute appropriately the situation of description take the behavior of proofreading and correct fast.

6. grind the method for the parts of system as one of one of above-mentioned claim described control cement mill, it is characterized in that: adopt the lab analysis data in (PCI) to refresh the fineness model parameter, condition is that the prediction of model is found and departs from the lab analysis value.

7. as the method for the parts of the system that grinds in one of above-mentioned claim described control cement mill, it is characterized in that: adopting currently provides (PCI) to determine whether interference (for example variation of feed fineness degree, humidity) is affecting process for operator and the employed intervention rule of slip-stick artist.

8. as the method for the parts of the system that grinds in one of above-mentioned claim described control cement mill, it is characterized in that: interfere information is made the behavior that is used to provide fast regular by (MPC) with feed-forward mode.

9. the control method of the parts of the system that grinds in cement mill described herein in conjunction with the accompanying drawings.

Images