Acoustic chemometric monitoring of an industrial granulation production process—a PAT feasibility study

Please download to get full document.

View again

of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Categories
Published
Acoustic chemometric monitoring of an industrial granulation production process—a PAT feasibility study
  Acoustic chemometric monitoring of an industrial granulation production process — a PAT feasibility study Maths Halstensen  a, ⁎ , Peter de Bakker   a,b , Kim H. Esbensen  a,c a   Applied Chemometrics Research Group (ACRG), Telemark University College/Tel-Tek, Department of Electrical Engineering, Information Technology and Cybernetics (EIC), Porsgrunn, Norway  b Yara, Sluiskil, The Netherlands c  ACABS (Applied Chemometrics, Analytical Chemistry, Applied Biotechnology, Bioenergy and Sampling) Research Group,University of Aalborg (AAUE), Esbjerg, Denmark  Received 5 January 2006; received in revised form 25 April 2006; accepted 1 May 2006Available online 5 July 2006 Abstract An acoustic chemometric Process Analytical Technologies (PAT) feasibility study of fluidized bed granulation of a fertilizer product (urea) isapplied to a semi-industrial pilot plant (SIPP) in a final evaluation before implementation for full-scale monitoring of the industrial production process. Conventional process monitoring and control is sub-optimal due to slow and labour-intensive laboratory analysis: particle size and liquidurea-concentration (correlated with water content) which are typically delayed by up to some 2+ h of analysis time. It is critical to be able to detect  process transients and unwanted upsets, e.g. critical conditions in the granulator which may lead to uncontrolled shutdown situations, necessitatingdays of labour-intensive work to clean the granulator. The acoustic chemometrics approach goes directly into a real-time, on-line domain. In thisstudy we focus on:1. Optimal localization of acoustic sensors2. Testing of a new sensor type (high-temperature microphone) in a semi-industrial granulator 3. Assessment of the feasibility to predict (by PLS-regression): •  bed movement (airflow trough granulator) •  liquid feed concentration (urea) •  reflux of fine material to granulator 4. Monitoring and visualization of critical trajectories – early warnings – in an operator-friendly fashion.Results show that both process state and product quality can be monitored to a satisfactory degree, e.g. detecting unwanted lump-formationalready at an incipient stage as well as bottom plate clogging. Such early warning allows process operators to change relevant process parameters(fluidization or atomization airflow, bed temperature, feed flux) to control product quality or to prevent critical shutdown situations. Successfulvalidation of this type of PLS-prediction model signifies that acoustic chemometrics is now maturing into a proven on-line technology in processanalytical chemometrics and PAT domains.© 2006 Elsevier B.V. All rights reserved.  Keywords:  Acoustic chemometrics; Process monitoring; Fluidized bed; Urea 1. Introduction This paper describes development of a new method tomonitor industrial granulation processes [1]. An earlier paper [2] described a first foray of small-scale pilot experiments, andgave an introduction to the acoustic chemometric approachspecifically in industrial granulation processes.General introduction to acoustic chemometrics has been published earlier  [3 – 5]; the following brief will suffice as background: acoustic chemometrics concerns capturing systemvibration characteristics, for example from two-phase systems Chemometrics and Intelligent Laboratory Systems 84 (2006) 88 – 97www.elsevier.com/locate/chemolab ⁎  Corresponding author.  E-mail address:  Maths.Halstensen@hit.no (M. Halstensen).0169-7439/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.chemolab.2006.05.012  (gas – solids/liquid – solids) generated e.g. by a manufacturing process or by transportation. The resulting vibrations can beeasily measured by non-intrusive,  “ clamp-on ”  sensors (accel-erometers). Acoustic signatures carry embedded informationabout a whole range of system-relevant physical and chemical parameters e.g. composition (oil, fat, ammonia, buttermilk,glycol, ethanol), mixing progress, fiber length, flow, density,temperature — as well as system state. For extraction andquantification of these types of specific analytes and parametersof interest, domain transforms (FFT, WT, AMT) and PLS-regression is essential for multivariate calibration.This approach is here used in order to obtain acousticmeasurements of   “ noise ”  produced by process equipment or  product movement in a semi-industrial granulator, used inthe experiments to produce a suite of specialized fertilizers;the present study focuses on urea exclusively. Other acousticmonitoring has earlier been explored on smaller fluidized bed processes within the same general context as here[9,10].Granulation of this kind of material is a complex process,which is controlled by experienced process operators. The parameters used to monitor the granulation process are so-calledstandard process measurements such as temperature, pressureand flow. The standard measurements have no information (or are only very indirectly related) to e.g. particle size, clogging of the reactor or the accumulating depository layering on the bottom plate — and often with a quite unacceptable delay time.A sample of layering cake on the bottom perforated plate takenout of the reactor after several days in production is shown inFig. 1, already a serious process impediment.When the layering cake develops further, the perforated bottom plate of the reactor necessarily becomes increasinglyclogged with a resultant fluidization airflow decrease.Decreased fluidization in turn leads to a situation with lessagitation of the particles; the result is often deformation of biglumps, which can quickly lead to a shutdown of the reactor, anda significant economic loss (reactor downtime and general production flow stoppage during clean-up).A measurement system, including chemometric models, that can predict the thickness of layering cake, particle size, or givean early warning of lump formation is thus highly wanted.Similarly, chemometric models for general process statemonitoring is of equally critical importance. Besides our conventional acoustic sensors (accelerometers [1 – 5], whichare all affixed on the outside of the reactor in a  “ clamp-on ” mode), a new sensor prototype was also tested out during the present experimental trials, viz. a high-temperature microphone probe inserted deep into the chamber. The objective of thisalternative sensor was to explore the possibility to have moredirect information about the particle formation/growth in thespraying zone near the reactor nozzles. The new prototypesensor was first inserted into one of the cooling chambers, to seeif it could withstand the harsh environmental conditions insidethe reactor. The intention was then to insert the sensor prototypeinto one of the spraying chambers if this first experiment wassuccessful.One of the major goals of the feasibility study is to relate process state trends, presented as chemometric score plots, tospecific conditions/qualities of the product inside the reactor.The process operators can then use this  “ new ”  informationtype to better operate the process, with an ultimate objectiveto significantly reduce (maybe even eliminate) costly shut-down situations. On-line measurements of particle character-istics such as particle size distribution together with propertiesof the liquid feed to the sprayer nozzles makes it manifestlyeasier to control the process. This paper concentrates on theresults from an experimental trial period of several months,involving a suite of induced deviations of the general production process in order to learn as much as possibleabout the feasibility of using acoustic chemometrics for theintended purposes.The objective of the present experiments was divided into 3major themes:1. Investigate different   sensor positions  on(in) a semi-industrialgranulator 2. Assessment of the  feasibility of acoustic chemometrics  to: •  predict bed movement (air flow through the granulator) •  predict concentration of the liquid feed •  predict reflux of fine material to the granulator  •  predict the moisture content in the granules •  predict average particle size and spread of the producedmaterial3. Monitor the overall granulator process state to detect   critical  situations  and to visualize these situations as early warningsin an operator-friendly fashion (lump formation and cloggingof the bottom plate are the most important mishaps in theindustrial production setting). 2. Experimental The experimental equipment consists of a semi-industrial pilot fluidized bed reactor, illustrated in Fig 2, which highlightsfive different sensor positions (A, B, C, D and the microphonelocation). All the first four sensors are mounted with screw-fittings onto a metal surface (in order to secure stable sensor  pick-up efficiency). Sensor position A is mounted onto an orifice plate  on the main supply line of liquid urea to thereactor nozzles, following Esbensen et al. [4]. Sensor positionsB, C and D are mounted directly onto the wall of reactor chambers 1, 2 and 4 respectively. All spectra used in all the Fig. 1. Layering cake on perforated fluidized bed bottom plate. Note extensiveclogging in one of the fluidization air holes with several others following suit.89  M. Halstensen et al. / Chemometrics and Intelligent Laboratory Systems 84 (2006) 88  –  97   Fig. 2. Semi-industrial pilot granulator (SIPP) used in all the experiments in this study. Sensor positions A, B, C and D are indicated. A microphone sensor was used inone experiment instead of sensor D, inserted into the first cooling chamber as shown.Fig. 3. PLS-1 model for airflow (m 3 /h) supply used to fluidize the particles in the granulator. Sensor position B was used. The model needed 8 PLS-components.Prediction results were validated using 10-segment cross validation. Conventional multivariate calibration:  “  predicted vs. measured ”  (top panel) and predicted vs. time(bottom panel). Note a few gross outliers which will be deleted in the final calibration.90  M. Halstensen et al. / Chemometrics and Intelligent Laboratory Systems 84 (2006) 88  –  97   chemometric models presented in this paper have 1024 data points (frequencies).The semi-industrial granulator displayed in Fig. 2 isidentical to an industrial full-sized granulator except for size, which is 1:10 roughly. The granulator is divided into fivechambers, three injection (spraying) chambers and twocooling chambers. The injection chambers each have severalnozzles where liquid urea is sprayed into the granulator at acertain process temperature. The bottom of the reactor is a perforated plate, which allows fluidization air to jet into thereactor, to interact with the growing particles and keep all particles in the bed in vigorous agitation. The cooling cham- bers are used to cool down the granules before they exit as thefinal product: urea granules with a specified size and sizerange (important parameters for agro-industrial product use).Sensor A is mounted onto an orifice plate inserted in themain supply pipeline for liquid urea. The orifice has a smaller hole diameter than the pipeline, which induces  turbulence  in theflowing urea downstream the orifice. The vibrations produced by this turbulence will be detected by sensor A. Further information of acoustic chemometrics and fluid flow in thiscontext can be found in [4]. Sensors B, C and D are mounted onthe vertical wall on the granulator, about 30 cm above the perforated bottom plate; they are supposed to detect   vibrations  produced by the granules when they interact with the reactor wall. Thus sensors B, C and D are used to monitor the processconditions inside the granulator, while sensor A is used to Fig. 4. PLS-1 model for urea melt concentration sprayed into the granulator in chambers 1, 2 and 3. Sensor Awas used in this model, which is based on 7 components.The model was validated with 10-segmented cross validation. Predicted vs. measured (top) and predicted vs. time (bottom). Note a few gross outliers which will bedeleted in the final calibration.91  M. Halstensen et al. / Chemometrics and Intelligent Laboratory Systems 84 (2006) 88  –  97   monitor the liquid supply of urea. The sensors used in this trialare four high-temperature accelerometers.The acoustic chemometrics experiments were not exclusivein the semi-industrial granulator context. The present measure-ments were in fact recorded in a  “  piggy-back  ”  mode, as other  process experiments – in themselves not related to acousticchemometrics – were carried out. This resulted in many dayswith stable conditions in the reactor, and no particular variationsin the acoustic signals. Therefore there were only a limitednumber of days (hours), which display the necessary  variationin process parameters , which are necessary for successfulmultivariate calibration. These still turned out to constitute asatisfactory basis for the full PAT feasibility study however. 3. Semi-industrial reactor experiments and results 3.1. Bed movement (sufficient particle agitation) One of the most important parameters in granulation processes concern  sufficient agitation  of the particles in thefluidized bed reactor. The overall agitation movement of  particles in the reactor has to be above a critical limit to prevent growing on the walls and/or lump formation. When thegranulator is working properly without any kind of lumps andlayering cake deposited on the bottom plate, the fluidizationairflow can be used as a reference to calibrate a model for the bed movement. When this model is used to predict the bed Fig. 5. PLS-1 model for reflux (kg/h) of fine material to chamber 1. Sensor B was used in this model, which is based on 7 PLS-components. The model was validatedwith 10-segmented cross validation. Predicted vs. measured (top panel) and predicted vs. time (bottom). Note a few gross outliers which will be deleted in the finalcalibration.92  M. Halstensen et al. / Chemometrics and Intelligent Laboratory Systems 84 (2006) 88  –  97 
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x