Predictive-Adaptive Temperature Control Of Molten Glass

2y ago
33 Views
2 Downloads
717.03 KB
5 Pages
Last View : Today
Last Download : 2m ago
Upload by : Jamie Paz
Transcription

Predictive-Adaptive Temperature Control of Molten GlassBill Gough, P. Eng.ANDUniversal Dynamics Technologies Inc.100 - 13700 International PlaceRichmond, BC V6V 2X8 Canadaemail: bgough@udtech.comAbstract - The temperature of molten glass is critical for theproduction of glass containers on automated molding machines.The temperature of the glass determines the quantity of glass inthe gob which is placed in the mold and thus affects the qualityof the finished container. There are frequent temperaturefluctuations in the glass as it exits the main furnace and threePID-controlled heat/cool sections are required to stabilize thetemperature before the gobs are cut for the molds. The longresponse time of these loops, combined with production ratechanges and the physical changes in the glass as a function oftemperature, make Proportional-Integral-Derivative controldifficult. This paper discusses the application of a newpredictive-adaptive controller for the control of glasstemperature. The new controller was able to reduce temperaturesettling times following set point or production rate changesfrom 4 to 6 hours to between 2 to 3 hours. This controlimprovement resulted in a significant reduction in rejectedcontainers as well as less maintenance and controller re-tuning.I. INTRODUCTIONThe production of glass containers presents many control andinstrumentation problems. Precise control of the molten glasstemperature is particularly difficult due to the changes in thephysical properties of the glass as a function of temperature.Typically, the glass temperature must be measured and controlledwithin 1 to 2 C out of a range of 1,100 to 1,300 C in order toproduce acceptable containers in the molding machine.The molten glass is produced in a large gas fired furnace fromsilica sand, soda ash, limestone, and other additives. The furnaceuses a combination of surface fired gas heating and immersedelectrode electric heating to melt the raw materials. Due to the poorthermal conductivity of the glass, the furnace operates in anDon MatovichConsumers Glass9622 Hill DriveLavington, BC V0E 2B0Canadaalternating gas fired-heat recuperation cycle to reduce fuel costs.This cycling introduces variations in the temperature of the glass asit exits the furnace and prevents the system from reaching a naturalsteady state.The molten glass is discharged from the furnace into a distributorwhere it flows into four separate forehearths. Each forehearth hasthree sections where the glass temperature is measured andcontrolled. The section closest to the furnace is called the RearSection, followed by the Front Section, then the ConditioningSection. The rear section and front section are combined heat/coolzones incorporating natural gas port valves and cooling windbutterfly valves operating inversely to bring the glass to the desiredtemperature. The last section is the conditioning zone, which isapproximately half the length of the two preceding sections and isequipped as a heating zone only (no provision for cooling). Themain function of these three zones is to provide a controlled,homogeneous cooling of the glass from the 1,500 C of the maintank (furnace) to the production temperature of 1,100 to 1,170 C.The glass pours out of an orifice in the Conditioning Section and issheared into discrete gobs. The gobs are guided through the air asthey drop by a series of automated chutes for delivery into theforming machine. A simplified diagram of the furnace andforehearths is shown in Fig. 1.The viscosity of the glass is very sensitive to temperature. If thetemperature changes, the amount of glass that pours through the gobcutter will change, affecting the resulting weight of the glasscontainer. Container weight is critical for proper molding in theforming machine. Correct container weight is essential to obtain thedesired container volume and appearance; thus it is a primaryquality parameter for the finished container.Furnace 2222-4RearFront22-1Conditioning22-322-2Gob CutterFig. 1. Simplified Diagram of Furnace and Forehearths.www.andritz.com1 of 5

II. EXISTING TEMPERATURE CONTROL SYSTEMThe existing control system used programmable ProportionalIntegral-Derivative (PID) controllers that were panel-mounted onthe factory floor. A PID controller was used to control the glasstemperature in each section of the forehearth. The controllers werealso connected to a graphical operator interface via a proprietarynetwork for data acquisition and trending.Control of the molten glass temperature is a difficult problem dueto: The response time of the temperature control loop is quite long(between 20 to 40 minutes);The combined heating/cooling control actuators are nonlinear;Production rate changes effect the gain and lag time for thetemperature control loop;Thermal and mechanical properties of the glass change withtemperature producing nonlinear dynamics.Operation of the forming machine involves both changes inproduction rate for a given container as well as changes in the typeof container to be produced (referred to as a “Job Change”).Following a production rate change or a Job Change, the glasstemperature would typically take between 4 to 6 hours to settle atthe set point temperature and enable the forming machine to producethe expected yield of acceptable containers (referred to as “StandardPack”). It was not uncommon for the glass temperature to require aslong as 10 hours to stabilize. In some cases, the glass temperaturecontinued to oscillate around set point for much longer periods andwould require several hours of attention by a knowledgeableinstrument technician to adjust the PID controller tuning andstabilize the process. Another problem was the operators becomingimpatient with the PID controllers and then switching to manualcontrol. This action often prolonged settling time because incorrectcontrol actions would be made by the operator.III. ADAPTIVE CONTROLA new predictive-adaptive controller was installed at the plant forcontrol of the molten glass temperature on forehearth 22-2. Thisforehearth was chosen because it was the most difficult to controldue to its near alignment with the furnace discharge throat. Thisforehearth was exposed to the largest swings in glass temperaturefrom the furnace because the glass spends the least time in thedistributor section, which is temperature controlled and tends tobuffer the temperature swings from the furnace.The adaptive controller was implemented on a PersonalComputer (PC) platform and was linked to the PID controllers usinga serial connection to the existing controller network. Glasstemperature, set point, and the control mode were read from the PIDcontrollers over the network. The PID controllers were configuredfor a “Tracking” control mode, which would allow the controlactions originating from the adaptive controller to be passed on tothe field actuators. This configuration provided the operators witheither Manual, PID, or Adaptive Control. This allowed the operatorsto continue to use a familiar interface and provided some security asthe existing control system was still available as a back-up to theadaptive controller. When the PID controller was switched toAdaptive Control, the PID algorithm was bypassed and the adaptivecontroller assumed control of the process.www.andritz.comThe adaptive controller is unique because of the technique usedto model the process. Dynamic Modeling Technology (DMT) is anew method of process transfer function modeling developed at theUniversity of British Columbia[1]. DMT reduces the effort requiredto obtain accurate process models. DMT is able to automaticallybuild a transfer function model using a series of orthonormalLaguerre functions. The Laguerre function series is defined as:li (t ) 2 pwhere:d i 1 i 1 2 pt t e (i 1) ! dt i 1 e pt(1)i 1 to Np Laguerre Polet timeA process transfer function can be approximated by summingeach function in the series where each function is multiplied by anappropriate coefficient or weighting factor:i g ( t ) ci li ( t )(2)i 0where:g(t) Process transfer functionc ith Laguerre coefficientThe DMT modeling method is able to represent higher orderprocess transfer functions and is inherently able to model processdead time. The user does not have to provide detailed knowledge ofthe process in order for an accurate transfer function to be obtainedresulting in a great reduction in the effort required to model theprocess. The DMT model is used as a basis for the design of thepredictive-adaptive regulatory controller.Using DMT, the adaptive controller is able to automatically adaptto changes in gain, time constants or time delay to maintain stablecontrol. The effects of measured process disturbances are alsomodeled in order to incorporate adaptive feed forward compensationinto the control strategy resulting in further performanceimprovements. The adaptive controller uses its mathematical modelsof the process to forecast process response so that set point isattained as rapidly as possible with little or no overshoot. The basicalgorithm steps used in the adaptive controller are shown in Fig. 2.The adaptive controller was installed to control glass temperaturein the Rear, Front, and Conditioning Sections of Forehearth 22-2.The temperature of the glass in the Rear Section was input as a feedforward for the Front Section temperature controller. Thetemperature of the glass in the Front Section was input as a feedforward for the Conditioning Section temperature controller. Thesefeed forward inputs allowed the adaptive controllers to anticipate thecontrol adjustments required to keep the glass temperatures at setpoint as the glass temperature in the preceding sections changed.2 of 5

compared to the existing PID controller’s performance. Theresulting effect of the glass temperature control performance on theyield of acceptable containers (pack) was then compared.A plot of the Rear, Front, and Conditioning Section temperaturesfollowing a pull change with the existing PID controllers is shownin Fig. 3 (this data was reproduced in essence from a paper stripchart recorder). The highest temperature is at the Rear Section andthe lowest temperature is at the Conditioning Section. Thetemperatures had not settled at set point after more than 5 hoursfollowing the pull change.Fig. 4 shows the performance of the adaptive controller followinga similar pull change that also involved a temperature set pointchange. The temperature stabilized in less than 3 hours, whichrepresents about a 50% improvement in temperature settling timecompared to the PID control system.Operating experience has shown that it is not unusual for the PIDcontrollers to take up to 6 hours or more to stabilize temperaturesfollowing a pull change. In some cases, the PID controllers wouldfail to stabilize the temperatures and the controllers would have tobe placed in manual mode until they could be re-tuned by aknowledgeable instrument technician. During this period, thetemperature control of the glass would be poor and the yield ofacceptable containers would be reduced. By comparison, theadaptive controllers have often been able to stabilize glasstemperatures in less than 2 hours and have not required adjustmentto maintain their performance.PROCESS LCULATECONTROLLEROUTPUTSET POINTINPUTPROCESSPLANTRESPONSEFig. 2. Basic Steps in the DMT Based Adaptive Controller.IV. CONTROL PERFORMANCE COMPARISONControl performance was evaluated based on the time requiredfor the glass temperature to stabilize following a production ratechange (referred to as a “Pull” change) which typically involves achange in the temperature set point. The ability of the adaptivecontroller to recover following a momentary gas shutoff was alsoPID Control Performance1220.00Temperature (Deg 9000.6000.3000.0001120.00Tim e (Hours)Fig. 3. PID Control Performance on a Pull Change.www.andritz.com3 of 5

8001.5001.2000.9000.6000.300Cond0.000Temperature (Deg C)AC Control PerformanceTim e (Hours)Fig. 4. Adaptive Control (AC) Performance on a Pull Change.PID-AC Control rature (Deg C)1220.001210.00Tim e (Hours)Fig. 5. PID/Adaptive Controller (AC) Comparison.Fig. 5 shows an example of a pull change where the PIDcontrollers were continuing to cycle for more than 8 hours after thepull change until the adaptive controller was enabled and was ableto stabilize the temperature. Later, the PID controller is re-enabledat 13 hours and the temperature again begins to cycle for more than12 hours resulting in a lower yield of acceptable containers. Notethat in this case the PID controllers would probably have to be retuned in order to be able to stabilize the temperatures.www.andritz.comForehearth 22-2 has about 20 job changes per month. Theimproved control with the adaptive controller saves approximately43 hours/month or 533 hours/year of lost production due to the glasstemperature not being stabilized at set point. This is about 6% of theannual production of forehearth 22-2.Occasionally, the gas supply to the forehearth is brieflyinterrupted causing a disturbance to the glass temperaturecontrollers. The PID controllers were observed to require about 1.54 of 5

hours to stabilize and the adaptive controllers required slightly lessthan 1 hour to stabilize glass temperature.The ultimate control performance comparison between theexisting PID controllers and the adaptive controllers is their effecton the production of acceptable glass containers. Over the last 2years of operating experience, the plant has observed animprovement in the pack 27% of standard pack for the mostcommon containers produced on forehearth 22-2. This represents aprofit increase of 533,820 per year with a resulting return oninvestment of less than 20 weeks. Occasionally, some specialtycontainers are produced which require more precise temperaturecontrol and as a result are particularly difficult to manufacture. Thepack for these containers has been observed to increase by as muchas 40 % with the adaptive controllers. The control comparisonresults are summarized in Table I.V. CONCLUSIONSControl of molten glass temperature is a difficult problem due tothe long response times, non-linear characteristics of the combinedheat/cool actuators, changes in the physical properties of glass as afunction of temperature, and changes in production rates. Theexisting PID temperature controllers required frequent re-tuning toprovide acceptable control of the glass temperature.Installation of an advanced predictive-adaptive controller hasdemonstrated that significant performance improvements could beachieved compared to the PID based control system. The timerequired for the glass temperature to settle following set pointchanges or production rate pull changes has been typically reducedby 50% with the adaptive controller. Resultant increases inproduction of on-spec containers of between 27% above standardpack have been achieved during the past 2 years for commoncontainers and production increases as high as 40% above standardpack have been observed for some specialty containers.The unique ability of the Dynamic Modeling Technology basedadaptive controller to learn the process and feed forward variablebehavior automatically and continuously ensures optimumperformance at all times. The problems of long development time,long setup time, repeated tuning and poor reliability associated withother advanced controllers such as Smith Predictor and other modelbased controller designs are solved with this method. The adaptivecontroller has resulted in a significant reduction in maintenance as itdoes not require the frequent re-tuning necessary for the PID basedcontrol system.The superior performance of this adaptive controller reducesprocess variability and enables the potential quality improvementbenefits of supervisory and statistical process control systems to berealized. In addition, the cascade effects of many smallimprovements provided by tighter control on individual loops canimprove the complete process or plant substantially. The DMTcontrol approach is a new tool available to the process controlengineer to economically implement the continuous improvementconcepts advocated by Deming [2] and Juran [3] in their TotalQuality philosophies.REFERENCES[1] Zervos, C.C., and Dumont, G.A., “Deterministic adaptivecontrol based on Laguerre series representation”, Int. J.Control, Vol. 48, No. 6, pp. 2333-2359, 1988.[2] Deming, W.E., “Out of the Crisis”, M.I.T. Center for AdvancedEngineering Studies, 1989.[3] Juran, J.M., “Juran’s Quality Control Handbook”, McGrawHill, 1988.TABLE I.CONTROL PERFORMANCE SUMMARYControl MetricPID ControlAdaptive ControlImprovementTemperature StabilityAfter Pull Change4 to 6 hours2 to 3 hours50%Temperature StabilityAfter Gas Interruption1.5 hours1 hour33%Pack (Common Containers)StandardStandard 27%27%Pack (Specialty Containers)StandardStandard 40%40%www.andritz.com5 of 5

predictive-adaptive controller for the control of glass temperature. The new controller was able to reduce temperature settling times following set point or production rate changes from 4 to 6 hours to between 2 to 3 hours. This control . Adaptive Control, the PID algorithm was bypassed and the adaptive

Related Documents:

The predictive cruise control (PCC) concept proposed in this brief utilizes the adaptive cruise control function in a predictive manner to simultaneously improve fuel economy and reduce signal wait time. The proposed predictive speed control mode differs from current adaptive cruise control systems in that be-

Sybase Adaptive Server Enterprise 11.9.x-12.5. DOCUMENT ID: 39995-01-1250-01 LAST REVISED: May 2002 . Adaptive Server Enterprise, Adaptive Server Enterprise Monitor, Adaptive Server Enterprise Replication, Adaptive Server Everywhere, Adaptive Se

predictive analytics and predictive models. Predictive analytics encompasses a variety of statistical techniques from predictive modelling, machine learning, and data mining that analyze current and historical facts to make predictions about future or otherwise unknown events. When most lay people discuss predictive analytics, they are usually .

Key–Words: Nonlinear Programming, Model Predictive Control, Receding Horizon Controller, Adaptive Control, Fixed Point Transformation 1 Introduction The classical realization of the Model Predictive Con-trollers (MPC) controllers [1, 2] applies the mathe-matical framework of Optimal Control (OC) in which

Adaptive Control, Self Tuning Regulator, System Identification, Neural Network, Neuro Control 1. Introduction The purpose of adaptive controllers is to adapt control law parameters of control law to the changes of the controlled system. Many types of adaptive controllers are known. In [1] the adaptive self-tuning LQ controller is described.

Adaptive Control - Landau, Lozano, M'Saad, Karimi Adaptive Control versus Conventional Feedback Control Conventional Feedback Control System Adaptive Control System Obj.: Monitoring of the "controlled" variables according to a certain IP for the case of known parameters Obj.: Monitoring of the performance (IP) of the control

Summer Adaptive Supercross 2012 - 5TH PLACE Winter Adaptive Boardercross 2011 - GOLD Winter Adaptive Snocross 2010 - GOLD Summer Adaptive Supercross 2010 - GOLD Winter Adaptive Snocross 2009 - SILVER Summer Adaptive Supercross 2003 - 2008 Compete in Pro Snocross UNIQUE AWARDS 2014 - TEN OUTSTANDING YOUNG AMERICANS Jaycees 2014 - TOP 20 FINALIST,

well as themes from PALS enquiries and formal complaints received within Western Sussex Hospitals NHS Trust during 2018. Patient experience monthly reports are provided to operational teams and patient comments are automatically shared with our staff. Leaders of our clinical services use the feedback we receive from patients to shape quality improvement activities at ward level and see whether .