WO2001006056A1

WO2001006056A1 - Method of estimating unknown or imprecisely known variables in a paper making process

Info

Publication number: WO2001006056A1
Application number: PCT/FI2000/000648
Authority: WO
Inventors: John Shakespeare; Jari HÄMÄLÄINEN; Petri Nyberg
Original assignee: Metso Paper Automation Oy
Priority date: 1999-07-15
Filing date: 2000-07-14
Publication date: 2001-01-25
Also published as: ATA91232000A; AU6284000A; CA2379228A1; DE10084794T1; AT410846B

Abstract

Method of estimating unknown or imprecisely known variables in a paper making process wherein a mathematical model of a process is formed, the model including at least one target variable which is unknown or imprecisely known. A plurality of process variables and at least one property of a manufactured paper are measured. At least one of said measured variables or properties is chosen as a reference variable. An initial value is chosen for the target variable for solving the mathematical model and a difference between the simulated reference variable and the measured reference variable is determined. The difference constitutes a cost function which is optimized to estimate the unknown or imprecisely known value of the target variable.

Description

METHOD OF ESTIMATING UNKNOWN OR IMPRECISELY KNOWN VARIABLES IN A PAPER MAKING PROCESS

The invention relates to a method of estimating unknown or imprecisely known variables in a paper making process.

US patent 5812404 discloses a method for the continuous overall regulation of a headbox of a paper machine. A physical fluid flow model of a pulp suspension flow discharged from the headbox to be regulated is formed and solved to obtain a simulated flow state based, e.g., on data on the geometry of the headbox and initial and boundary conditions related to the headbox. A target flow state based on the quality requirements of the paper produced from the pulp suspension flow, the costs of operation and runnability of the machine is obtained and a difference between the simulated flow state obtained as the solution of the flow model and the target flow state is determined. This difference constitutes a cost function which is optimized to determine optimal regulation values and set values for instrumentation devices and actuators of the headbox which affect the pulp suspension flow.

Provision of accurate initial and boundary conditions is advantageous in a method disclosed in US 5812404 but the solution of US 5821 404 relies on measurements alone to provide said conditions. Measurements alone cannot provide all required initial values with adequate accuracy. The values of variables are also too imprecise or unknown at several other points of the paper making process, and thus it would be necessary to obtain a more accurate estimate of the variable values. An object of the present invention is to provide a method for obtaining a more accurate estimate of unknown or imprecisely known variables of a paper making process.

The method of the invention is characterized by comprising the steps of forming a mathematical model of a process, the model including at least one target variable which is unknown or imprecisely known, measuring a plurality of process variables and at least one property of a manufactured paper web, utilizing the measurements as initial and boundary values of the model, choosing at least one of said measured process variables or measured web properties as a reference variable, choosing an initial value for the target variable, solving the mathematical model, simulating a value of the reference variable utilizing the model, determining a difference between the simulated reference variable and the measured reference variable, the difference constituting a cost function, and optimizing the cost function to estimate the unknown or imprecisely known value of the target variable.

The essential idea of the invention is that a mathematical model of a process is formed, the model including at least one target variable which is unknown or imprecisely known. Plural process variables are measured and at least one property of the manufactured paper web is measured. At least one of said measured process variables or measured web properties is chosen as a reference variable. Boundary conditions for the mathematical model are obtained from the remaining measured process variables and measured web properties. An initial value is chosen for the target variable for solving the mathematical model, the reference variable is thereby simulated, and a difference between the simulated reference variable and the measured reference variable is determined. This difference constitutes a cost function which is optimized to estimate the unknown or imprecisely known value of the target variable. The idea of a preferred embodiment is that the process is a flow process in a headbox and the target variable is a variable of the headbox. An advantage of the invention is that variables of the paper making process can be estimated in a more accurate and diverse manner than earlier, which allows to improve control of the paper making process. The method applies particularly to estimation of unknown or imprecisely known variables which are multivariate or vector quantities, such as the cross machine profile of a process variable or web property, or the statistical distribution of a variable whose value changes with position or time.

In this disclosure the term "paper" also refers to paper board and tissue.

The invention will be described in greater detail in the accompanying drawings, in which

Figure 1 is a schematic side view of a paper making process, Figure 2 is a simplified block diagram of a method of the invention, and

Figure 3 is a simplified flowchart for an optimizing controller in which the method of the invention can be utilized. Figure 1 schematically shows a paper machine. The paper machine comprises a headbox 1 , from which pulp is fed into a former 2, where a fiber web 3 is formed of the pulp. The web 3 is conveyed to a press 4 and further to a dryer unit 5. From the dryer unit 5 the web 3 is conveyed to a reel 6. A paper machine may also comprise, for example, a size press, coaters or a calender, which are not illustrated in the Figure 1 for the sake of clarity. Furthermore, the function of the paper machine is known per se for those skilled in the art, and will therefore not be further explained in this context.

The paper machine also comprises at least one measuring device 7 for measuring properties of the paper web 3. The measuring device 7 is used for measuring e.g. the basis weight BW, dry weight profile DWP and the fiber orientation angle profile FO of the paper web 3. The paper machine also comprises other measuring devices for measuring values and properties at different parts of the paper machine. Headbox pressure P, headbox consistency Cs, a profile of dilution valve positions DVP and headbox upper lip profile ULP are given as examples of such measurements in Figure 1. Additional properties of the paper web or of process streams can be measured from samples taken from the manufactured paper or process streams using instruments located elsewhere.

The invention relates to a method for estimating unobserved variables or imprecisely measured process variables in a paper machine during operation, and especially for estimating an unknown or imprecisely measured process variable in the headbox 1 of a paper machine or in the sheet formed of the headbox discharge. A mathematical model of the process is formed based for example on the geometry of the headbox 1 and physical models of the flow phenomena occurring therein, and including at least one target variable which is unknown or imprecisely known. Plural process variables are measured, for example using instruments in or around the headbox 1 and the various process streams leading to or from the headbox, and at least one property of the manufactured paper web 3 is measured using instruments elsewhere on the machine or in a laboratory. At least one of said measured process variables or measured web properties is chosen as a reference variable. Boundary conditions for the mathematical model are obtained from the remaining measured process variables and measured web properties. An initial value is chosen for the target variable for solving the mathematical model, the reference variable is thereby simulated, and a difference between the simulated reference variable and the measured reference variable is determined. This difference constitutes a cost function which is optimized to estimate the unknown or imprecisely known value of the target variable. The initial value chosen can be, for example, an imprecise measurement of the target variable, so that the optimization produces an estimate which is of superior accuracy to the measurement.

A simplified flow chart for the estimation process is depicted in Figure 2. The physical and geometric parameters are known a priori. The measurements of known states are provided by measuring instruments or operator input. The initial estimate of the unknown variable or state may be supplied from an approximate measurement, or may be provided by other means of estimation, or by an operator.

For example, in the headbox 1 the slice opening profile is approximately known. The profile of the upper lip ULP is measured at a number of locations across the slice, but the shape of the lower lip is generally unknown. However, various profiles which depend on the slice opening profile are measured, such as the basis weight profile BW or dry weight profile DWP and the fiber orientation angle profile FO.

The optimization uses the measured slice lip shape as an initial estimate of the slice opening profile. One or more suitable measurements, such as the fiber orientation angle profile FO or dry weight profile DWP or basis weight profile BW are used as reference variables. Other measurements, such as headbox pressure P, consistency Cs, temperatures, flows, a profile of dilution valve positions DVP etc. are used as inputs to a mathematical model of the flow process in the headbox 1 and jet. By solving the mathematical model of the headbox 1 , a simulation of the fiber orientation angle profile is obtained. A penalty to be optimized is calculated from the difference between the measured fiber orientation angle profile and simulated fiber orientation angle profile. This penalty is optimized by adjusting the estimate of the slice opening profile, the optimum being reached when the simulated fiber orientation angle profile most closely matches the measured fiber orientation angle profile. This invention solves a mathematical model of the headbox to estimate unknown or imprecisely known current operating parameters consistent with measured properties of the current product.

The present invention is particularly well suited for estimation of the initial conditions for a method of optimizing control such as that disclosed in US 5812404 which is incorporated herein by reference. The present invention uses inverse modeling to provide such a controller with improved estimates of its initial and boundary conditions. The present invention can utilize similar mathematical and physical models to those used by such a controller, or can use different ones.

A simplified flow chart for an optimizing controller is depicted in Figure 3. One or more of the "measurements of known variables/states" in that figure can be supplied by the estimation process of this invention.

The drawings and the description thereof are merely intended to illustrate the inventive idea. The details of the invention may vary within the scope of the claims. Thus the method of the invention can in principle be applied at any point of the paper making process.

Claims

1. A method of estimating unknown or imprecisely known variables in a paper making process, the method comprising the steps of forming a mathematical model of a process, the model including at least one target variable which is unknown or imprecisely known, measuring a plurality of process variables and at least one property of a manufactured paper web, utilizing the measurements as initial and boundary values of the model, choosing at least one of said measured process variables or measured web properties as a reference variable, choosing an initial value for the target variable, solving the mathematical model, simulating a value of the reference variable utilizing the model, determining a difference between the simulated reference variable and the measured reference variable, the difference constituting a cost function, and optimizing the cost function to estimate the unknown or imprecisely known value of the target variable.

2. A method according to claim 1 , wherein the process is a flow process in a headbox and the target variable is a variable of the headbox.

3. A method according to claim 2, wherein the target variable is a slice opening profile of the headbox.

4. A method according to claim 3, wherein the reference variable is a fiber orientation angle profile.

5. A method according to claim 3, wherein the reference variable is a basis weight profile.

6. A method according to claim 3, wherein the reference variable is a dry weight profile.

7. A method according to claim 2, wherein the measured process variables comprise headbox pressure and headbox consistency.

8. A method according to claim 2, wherein the measured process variables comprise a profile of dilution valve positions.