Çimento endüstrisinde harmanlama prosesinin öz uyarlamalı kontrolü
Başlık çevirisi mevcut değil.
- Tez No: 46144
- Danışmanlar: PROF.DR. CAN ÖZSOY
- Tez Türü: Yüksek Lisans
- Konular: Makine Mühendisliği, Mechanical Engineering
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 1995
- Dil: Türkçe
- Üniversite: İstanbul Teknik Üniversitesi
- Enstitü: Fen Bilimleri Enstitüsü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 90
Özet
with the use of a forgetting factor (3, which is slightly less than unity. By using this, we obtain exponential forgetting of the past data. Instead of giving equal eight to the errors in the RLS, we give more weight to more recent data. It rages from 0.95 to 0.999. In this thesis, multiple input multiple output minimum-variance self-tuning algorithm was experienced on raw meal blending in cement industry. The control of the chemical composition of the ground mix of the raw materials and its homogenization before feeding it into the kiln is a very essential problem in cement manufacture. The aim of the blending control system is to produce a full silo of kiln feed at the desired chemical composition with minimum variation through the silo filling. The control problem arises from the fact that the chemical compositions of the various raw materials vary from time to time because of geographical site changes in quarry and the time delay between obtaining the sample and making any correction to the raw mill inputs can be several hours in spite of x-ray fluorescence technique (XRF) applied for analyzing the chemical composition of the raw meal. XRF analyzer require a high degree of preliminary sampling and are ineffective for frequent raw materials analysis. One of the major problems in designing raw materials blending control is the difficulty of obtaining frequent rapid accurate representative chemical analyses of the raw materials. This problem can be highly eliminated by prompt gamma neutron activation analysis (PGNAA). PGNAA provides continuous on-line raw material blending control. The weight feeders are controlled by computer. The meal is fed into the mill by a conveyor belt. The raw meal which is made up of raw materials of different compositions are sampled before the silo. The oxides compositions obtained by a x- ray fluorescence analyzer (RFA) are provided for the computer which calculates the new values to give to each feed flow with regard to a fixed total feed flow for the weight feeders. The primary purpose of the blending control is to reduce the feed composition disturbances to the kiln and improve the quality control. There are four most important oxides: S(Si02), A(A1203), F(Fe203), C(CaO). These the most important oxides or from them the following four moduli are computed by the computer: Lime moduli ML=100C/2.8+l.lA+0.8F Aluminum moduli MA=A/F Silica moduli MS=S/A+F Hydraulic moduli MH=C/S+A+F
Özet (Çeviri)
with the use of a forgetting factor (3, which is slightly less than unity. By using this, we obtain exponential forgetting of the past data. Instead of giving equal eight to the errors in the RLS, we give more weight to more recent data. It rages from 0.95 to 0.999. In this thesis, multiple input multiple output minimum-variance self-tuning algorithm was experienced on raw meal blending in cement industry. The control of the chemical composition of the ground mix of the raw materials and its homogenization before feeding it into the kiln is a very essential problem in cement manufacture. The aim of the blending control system is to produce a full silo of kiln feed at the desired chemical composition with minimum variation through the silo filling. The control problem arises from the fact that the chemical compositions of the various raw materials vary from time to time because of geographical site changes in quarry and the time delay between obtaining the sample and making any correction to the raw mill inputs can be several hours in spite of x-ray fluorescence technique (XRF) applied for analyzing the chemical composition of the raw meal. XRF analyzer require a high degree of preliminary sampling and are ineffective for frequent raw materials analysis. One of the major problems in designing raw materials blending control is the difficulty of obtaining frequent rapid accurate representative chemical analyses of the raw materials. This problem can be highly eliminated by prompt gamma neutron activation analysis (PGNAA). PGNAA provides continuous on-line raw material blending control. The weight feeders are controlled by computer. The meal is fed into the mill by a conveyor belt. The raw meal which is made up of raw materials of different compositions are sampled before the silo. The oxides compositions obtained by a x- ray fluorescence analyzer (RFA) are provided for the computer which calculates the new values to give to each feed flow with regard to a fixed total feed flow for the weight feeders. The primary purpose of the blending control is to reduce the feed composition disturbances to the kiln and improve the quality control. There are four most important oxides: S(Si02), A(A1203), F(Fe203), C(CaO). These the most important oxides or from them the following four moduli are computed by the computer: Lime moduli ML=100C/2.8+l.lA+0.8F Aluminum moduli MA=A/F Silica moduli MS=S/A+F Hydraulic moduli MH=C/S+A+FÖZET Bu tezde, çok girişli çok çıkışlı minimum varyans öz uyarlamak regülatörlerin (Multiple Input Multiple Output Minimum Variance Self-Tuning Regulators veya MTMO- MV-STR) özellikleri, çeşitleri ve kullamm sahaları araştırılmış ve gerçek bir çimento tesisinde hammadde harmanlanması prosesi üzerinde denenmiştir. Ayrıca çok girişli çok çıkışlı minumun varyans öz uyarlamak regülatörlerin daha yüksek verimde çalışması amacına hizmet eden sonlu zamanda gerekli ortalama kriteri ( Required Average For Finite Time veya RAFT) aynı proses üzerinde incelenmiştir. Harmanlama kontrolundan amaçlanan, firm beslemesinin istenilen kimyasal kompozisyonunun sağlanmasıdır. Silo dolumu sırasında, ham maddenin kimyasal kompozisyonunda minimum değişim istenir. Kontrol problemi, çeşitli hammaddelerin kimyasal kompozisyonlarının zamanla değişmesi ve bu değişimlerin anında analiz edilip gerekli değişikliğin yapılamamasından kaynaklanmaktadır. Bu nedenle; harmanlama prosesinde öz uyarlaman kontrol için gerekli kimyasal analiz yöntemleri olan Hızlı Gama Nötron Aktivasyon (Prompt Gamma Neutron Activation Analysis veya PGNAA) ve X- Işını Fluoresans analiz yöntemleri (X-Ray Fluorescence Analysis veya XRF) de araştırılarak birbirine göre üstünlükleri belirlemniştir. Ham madde harmanlama prosesi, besleme bunkerlerinde bulunan farklı kompozisyonlardaki ham maddelerin, bilgisayarla kontrol edilen tartılara dökülmesiyle başlar ve istenilen hammadde ağırlıkları sağlandıktan sonra bir konveyörle değirmene beslenir. Değirmen çıkışında belirli periyotlarla örneklenen harman, bilgisayar tarafından kullanılacak olan oksit analizleri için x-ışım analiz cihazına gönderilir. Daha sonra ortalama değer etrafındaki sapmaların azaltılması için hava ile homojenleştirme işlemine tabi tutulan dolum silolarına, döner finnı beslemek üzere sevk edilir. Bu projenin sağlayacağı yararlar: I. Serbest kireç değişiminin manuel kontrola göre 1/4 kadar daha azaltılması II. Yakıt sarfiyatında %4 kadar azalma IH. Ürün kalitesinde artış rv. Üretimde artış V. Maliyetin düşürülmesi VI. Çalışma düzenlihğinde artış VII. Enerji ve çevre yönünde kazançwith the use of a forgetting factor (3, which is slightly less than unity. By using this, we obtain exponential forgetting of the past data. Instead of giving equal eight to the errors in the RLS, we give more weight to more recent data. It rages from 0.95 to 0.999. In this thesis, multiple input multiple output minimum-variance self-tuning algorithm was experienced on raw meal blending in cement industry. The control of the chemical composition of the ground mix of the raw materials and its homogenization before feeding it into the kiln is a very essential problem in cement manufacture. The aim of the blending control system is to produce a full silo of kiln feed at the desired chemical composition with minimum variation through the silo filling. The control problem arises from the fact that the chemical compositions of the various raw materials vary from time to time because of geographical site changes in quarry and the time delay between obtaining the sample and making any correction to the raw mill inputs can be several hours in spite of x-ray fluorescence technique (XRF) applied for analyzing the chemical composition of the raw meal. XRF analyzer require a high degree of preliminary sampling and are ineffective for frequent raw materials analysis. One of the major problems in designing raw materials blending control is the difficulty of obtaining frequent rapid accurate representative chemical analyses of the raw materials. This problem can be highly eliminated by prompt gamma neutron activation analysis (PGNAA). PGNAA provides continuous on-line raw material blending control. The weight feeders are controlled by computer. The meal is fed into the mill by a conveyor belt. The raw meal which is made up of raw materials of different compositions are sampled before the silo. The oxides compositions obtained by a x- ray fluorescence analyzer (RFA) are provided for the computer which calculates the new values to give to each feed flow with regard to a fixed total feed flow for the weight feeders. The primary purpose of the blending control is to reduce the feed composition disturbances to the kiln and improve the quality control. There are four most important oxides: S(Si02), A(A1203), F(Fe203), C(CaO). These the most important oxides or from them the following four moduli are computed by the computer: Lime moduli ML=100C/2.8+l.lA+0.8F Aluminum moduli MA=A/F Silica moduli MS=S/A+F Hydraulic moduli MH=C/S+A+Fadvent of microprocessors, which make implementation of STR algorithm relatively easy. A self tuning controller structure is shown in fig. 1. Fig.1: Block Diagram of a Self Tuning Regulator The aim of self-tuning and adaptive systems is to automate in some way certain activities of the control system. The principle tasks involved in control system consists of the following: I. modeling of a system II. design of a controller ffl.implementation of the controller These three stages of the control system is shown in fig. 2. In principle, there are two different ways in which models can be obtained: from prior knowledge -e.g., in term of physical laws-or by experimentation on a process. In most cases it is not possible to make a complete model only from physical knowledge. Some parameters must be determined from experiments. This approach is called system identification. There are many methods for analyzing data obtained from experiments. One basic approach is the principle of least squares and recursive ways to make the computations. A model is very useful and compact way to summarize the knowledge about a process. The process models can sometimes be obtained from first principle of physics. It is more difficult to get the models of the disturbances, which are equally important. These models often have to be obtained from experiments. Experiments are often the only way to get models for the disturbances. A process should be represented by a hierarchy of models ranging from detailed and complex simulation models to very simple models, which are easy to manipulate analytically. The simple models are used for exploratory purposes and the complicated models.are used for a detailed check of the performance of the control system. There are no general method that always can be used to get a complete model Each process or problem has its own characteristics. The main problem when making a mathematical model is VIIto find the states of the system. The state variables essentially describe storage of energy and mass in the system. The advantage of model-building from physics is that it gives insight; also, the different parameters and variables have physical interpretations. The drawback is that it may be difficult and time-consuming to build the model from first principles. Mathematical model-building often has to be combined with experiments. Mathematical Presentation of a System Modeling ) - /"^DesigrT Satisfactory _ ___ ^?^?1^^ >v ( Objective Unsatisfactory Controller Fig.2. : The Three Stages of Control System System identification is the experimental approach to process-modeling. System identification includes the following: I. Experimental planning II. Selection of model structure m.Parameter estimation revalidation It is often difficult and costly to experiment with industrial processes. Therefore, it is desirable to have identification methods that do not require special input signals. A good identification method should be insensitive to the characteristics of the input signal It is sometimes possible to base system identification on data obtained under closed- loop control of the process. This is useful from the point of view of applications. The model structures are derived from prior knowledge of the process and the disturbances. When formulating an identification problem, a criterion is introduced to give a measure of how well a model fits the experimental data. The criteria for discrete-time systems are often expressed as the sum of squares of prediction error. The first formulation, solution and application were given by Gauss. Gauss formulated the identification problem as an optimization problem and introduced the vmIn this thesis, first the most important four oxides were chosen as outputs and required raw materials in percentage for cement; clay, lime, iron ore, and kaolin were as inputs. Second from the most important four oxides, the modules; ML, MA, MS, and MH were chosen as outputs even if the inputs were the same. It was observed that in the case of second choice, the sum of measured inputs in percentage was more close to 1. This means that the second choice is more suitable for raw meal blending control. It is necessary, not only to minimize the variance of the outputs but it is very desirable to keep its weighted average for finite N as close to the reference as possible. To satisfy this demand self-tuning strategy using time varying reference value was experienced. The required average for finite time (RAFT) strategy meets the following requirements: The general objective is to control the average composition of a full silo according to the chemist's reference values and control the instantaneous mill output composition y(t) as closely as possible to the reference vector, but still allowing correction for the unavoidable past deviations. This combination of silo and mill composition control provides the desired kiln feed composition and minimizes blending ( homogenizing ) requirements. Thus at the end of silo feeding the average composition must reach the given reference value at the N Th. interval and the variation of y(t) around the reference has to be minimized. Results reveals the following improvements: I. Raw material saving because of reduced purchasing of additives II. Reducing kiln accretions HI. Reduced consumption of ammunition, ie. less annular accretions IV. Shorter downtimes of the kiln V. Reduced specific energy consumption of the cement mill thank to more uniform clinker quality XI
Benzer Tezler
- Çimento endüstrisinde hammadde harmanlama prosesinin klasik ve akıllı yöntemler ile modellenmesi
Modling of raw material blending process in cement industry using conventional and intelligent techniques
KEMAL KIZILASLAN
Yüksek Lisans
Türkçe
2004
Makine Mühendisliğiİstanbul Teknik ÜniversitesiMakine Mühendisliği Ana Bilim Dalı
Y.DOÇ.DR. ŞENİZ ERTUĞRUL
- Çimento endüstrisinde hammadde harmanlama prosesinin optimal kontrolü
Optimal control of raw material blending process in cement industry
GÜRKAN TUTUM
Yüksek Lisans
Türkçe
1997
Makine MühendisliğiSakarya ÜniversitesiMakine Mühendisliği Ana Bilim Dalı
YRD. DOÇ. DR. RECEP KAZAN
- Indentification and model predictive control of the raw material blending process in cement industry
Çimento endüstrisinde hammadde harmanlama prosesinin tanılanması ve model öngörülü kontrolü
AYHAN KURAL
- Çimento endüstrisindeki harmanlama projesinin değişik tipte yapay sinir ağları ile tanılanması ve kıyaslanması
The Identification and comparison of the blending process in cement industry with various type of neural networks
GÜVEN EMRE
Yüksek Lisans
Türkçe
2003
Makine Mühendisliğiİstanbul Teknik ÜniversitesiMakine Mühendisliği Ana Bilim Dalı
PROF. DR. CAN ÖZSOY
