Algorithms for Raw Material Dosage Control for Cement Plants

Transcription

1 Algorithms for Raw Material Dosage Control for Cement Plants STELA RUSU-ANGHEL, OVIDIU TIRIAN, OSACI MIHAELA, DINIȘ CORINA Department of Electrotechnical Engineering and Industrial Informatics Politechnica University of Timisoara Revolutiei str., no. 5, Hunedoara ROMANIA {stela.anghel, ovidiu.tirian, osaci.mihaela, Abstract: - The raw material used for producing cement is obtained using several components which we must dosages so that the mixture should have a uniform chemical composition. This paper work proposes fuzzy algorithms for performing this process. Using them turns the control dosage system more simple and it allows any high performance. The system we propose has been tested in industrial conditions and the results have backed up any hypothesis. Key-Words: dosage, raw material, cement, fuzzy, control, algorithm. 1. Introduction In order to produce cement, we use the following raw material limestone, clay, and pyrite ash. There are other receipts as well, depending on the specific of any area. Depending on the total quantity we need and on chemical composition of each, all the components are dosage so that the mixture had the same required and uniform chemical composition. After the mixture has been prepared, it is milled and conditioned to obtain a granulation and moisture requirements and it is put into the rotative furnace. Inside the furnace there are some specific and complex phenomena which produce clinker. Finally, the clinker is ground and combined with other additives formed along the cement. This paper tries to come up with a control system (hardware-software) for the raw material dosage for any cement plant, using fuzzy logics. If we use this principle it allows us to keep the chemical composition as constant as possible, in order to produce quality cement, without any composition variation. 2. System structure and control algorithm Goal Maintaining the required modulus (GSMS and MAlM) [7] of the reference values by adjusting the proportions of limestone, clay, and pyrite ash at feeding of the mill. The proposed structure of the system for automatic dosage of raw material is represented in figure 1. Figure 1. Automatic raw material dosage structure. ISSN: ISBN:

2 This process is led by a schedule automatic device (AP) with input/output modulus and connected to a network belonging to a higher level. We must use some intelligent transducers which are able to communicate with the schedule automatic device through a local network. AP contains a fuzzy control device [2], [3], whose algorithm is presented into the following paragraph. The rest of the equipment we use in this scheme is typified [1], [4], [5], [6]. Using fuzzy algorithms need a periodical analysis of specific chemical (oxides) composition values for the limestone, the clay, and the pyrite ash (they are compulsory in case of the determining algorithm). For adjustment of the proportions as close the reality, we should analyze the raw materials statistically. Algorithm stages: a) Reading the limestone flow (DC), clay flow (DA), and pyrite ash flow (DP), from process. There are gravimetric dosage devices for the raw materials. We must connect the three analogical (flow) values for a mill. b) Sampling and analysis of samples of flour (mill + electro-filter) after almost 1 hour the sampling interval. Required: device for sampling flour; an X-ray analyzing device for determining the oxide composition of the flour (CaO, SiO 2, Al 2 O 3, Fe 2 O 3, etc.) and for calculating the modulus (GSO, MAl, MSi), in case of new analyzing devices; contact for reporting the time of sampling (two numerical inputs); total calculation sample-sample time and the flow input time. c) Calculation of feed mill flour with limestone, clay, and pyrite ash between two samplings. This should be made softly, by integrating the feeding DC, DA, DP. Flour mill production (QMF) (equivalent of the quantity of the flour introduced into the homogenization silo): QMF e = QC + QA + QP (1) ΣDC ΣDA ΣDP d) Calculation of real proportions in feed of the flour mill between two sample-sample periods (soft): Limestone percentage: QC PRC= 100 % (2) QMFe Clay percentage: QA PRA= 100 % (3) QMF e Pyrite ash percentage: QP PRP= 100 % (4) QMFe e) Calculation of the modulus composition of flour between sample-sample intervals: GSO e, MSi e, MAl e [7]: %CaO GSO e = (5) 0.8%SiO + 1.1%Al O + 0.7%Fe O 2 %SiO 2 MSi e = (6) %Al O + %Fe O %Al O MAl e = (7) %Fe O It is comprised by fuzzy dosing algorithm - GSO, MSi, MAl. The oxides values are made by X- ray analysis. Observation: There are some analyzing devices which record the modulus values directly. f) Calculation of the entire flour quantity inside the silo. This is made softly, by adding the flour quantities inside the homogenization silo after each stage. n QMF T = QMF e (8) e= 1 when n is the number of dosage stages (samplesample inputs). g) Calculation of the composition (oxides and modulus) of the flour inside the homogenization silo (corresponding to QMF T ) after each dosage stage (sample-sample inputs). It is softly made, by weight control. For instance, in order to calculate the composing of CaO of flour inside the silo, we use the relation (for n stage): (CaO T ) n =%(CaO T ). n-1 (QMF T ) n-1 + for the flour inside the silo up to n stage +%(CaO e ). n (QMF e ) n (9) for the flour inside the silo at n-stage The final calculus is (GSO T ) n, (MSi T ) n, and (MAl T ) n, where there are GSMS, MSiM, and MAlM, in case of a n-stage, and which comprises the fuzzy dosage algorithm. h) Calculation of the adjusted percentage of limestone and pyrite ash, according to the fuzzy algorithm. For limestone: PRC n = PRC n-1 + ΔPRC n (10) For pyrite: PRP n = PRP n-1 + ΔPRP n (11) i) Calculation of the adjustment for the clay percentage: PRA n = ( PRC n + PRP n ) (12) j) Transmission of the new percentages to the ISSN: ISBN:

3 process: PRC n, PRA n, PRP n. For any new sampling stage, the cycle follows the same pattern: - sampling; - determining the percentage (%) of the oxides and flour modulus; - using fuzzy algorithm coefficients (ΔPRC n, ΔPRP n ); - calculation of the new percentage of limestone and pyrite ash; - calculation the new percentage of clay; - new percentages (inputs values). Any former percentage becomes a reference value. In order to perform any percentage (limestone, clay, pyrite ash) we need to couple at the calculation system for eight numerical outputs (flow operations). 3. Fuzzy control algorithm 3.1. Input size information a) MAl (Aluminum modulus/hour) Number of states MAl = 3 S LOW N NORMAL R HIGH Figure 3. Memberships functions for MAlM. c) GSO (saturation degree/hour) Number of states GSO = 5 FS VERY LOW 0 92 S LOW N NORMAL R HIGH FR VERY HIGH Figure 4. Memberships functions for GSO. d) GSMS (average saturation degree/silo) Number of states GSMS = 3 S LOW 0 96 N NORMAL R HIGH Figure 2. Memberships functions for MAl, using fuzzy control device software. b) MAlM (average Aluminum modulus/silo) Number of states MAlM = 3 S LOW N NORMAL R HIGH Figure 5. Memberships functions for GSMS Control values informations a) dprp (percentage adjustment for pyrite ash) ISSN: ISBN:

4 CM HIGH INCREASE +1.5 CP LOW INCREASE M LEVELLED 0 SP LOW DECREASE SM HIGH DECREASE -1.5 Figure 8. Inference table for dprp. Figure 6. Memberships functions for dprp. b) dprc (limestone percentage adjustment) CFM ERY HIGH INCREAS CM HIGH INCREASE +1.5 CP LOW INCREASE M LEVELLED 0 SP LOW DECREASE SM HIGH DECREASE -1.5 SFM ERY HIGH DECREAS Figure 9. Inference table for dprc. Figures 10 and 11 represent the stimulation of the fuzzy control. Figure 7. Memberships functions for dprc Control rules a) for MAl from figure 8 Total number of rules MAl = (Number of states MAl)*(Number of states MAlM)=9 b) For GSO from figure 9 Total rule number GSO = (Number of states GSO)*(Number of states GSMS)=15 Figure 10. Simulation of the fuzzy control MAI. ISSN: ISBN:

5 4.2 Limestone percentage adjustment (dprc) Input size: GSO (saturation degree/hour) and GSMS (average saturation degree/silo). Uni-polar domain (0 100%) for GSO and GSMS, and bipolar for dprc. Control surface dprc=f(gso, GSMS), for the Fuzzy controller, is presented in figure 13: Figure 11. Simulation of the fuzzy control GSO. 4. Static maps for proposed controllers 4.1 Pyrite percentage adjustment (dprp) Input values: MAl (Aluminium modulus/hour) and MAlM (Average Aluminium modulus/silo). Uni-polar domain (0 100%) for MAl and MAlM, and bi-polar for dprp ( %). Control surface dprp=f(mal, MAlM) for the Fuzzy controller is represented in figure 12. Figure 12. Control surface for dprp. It is not very difficult to read those features. They correspond entirely to the inference table for dprp (two inputs, one output and 9 rules). Figure 13. Control surface for dprc. 5. Conclusions This paper tries to come up with a control system (hardware-software) for the raw material dosage for any cement plant, using fuzzy logics. If we use this principle it allows us to keep the chemical composition as constant as possible, in order to produce quality cement, without any composition variation. Besides the classical methods, the algorithm we propose is simple, it does not need any expensive hardware, and it is easy to implement it to any cement plant. All research are original and have been tested in a Romanian plant. The results of the test have confirmed that the algorithms had been correct [9], [10]. References: [1] Anghel,S. ş.a., Expert Raw Material Dosage Management for Cement Production, based on Fuzzy Logics, Jubilee Scientific Communication Conference, Reşiţa, October 25th-26th, 1996, Vol.3, pp [2] Anghel,S., Universal Fuzzy Control Device for Industrial Process Management - hardware, Scientific Communication Confference, Hunedoara, Octomber 31st - November 1st. 1997, Vol.3, pp [3] Anghel,S., Universal Fuzzy Control Device Software, Scientific Communication Confference, Hunedoara, October 31st - November 1st. 1997, Vol.3, pag [4] Brown-Boveri (BBC), Automation Packages for ISSN: ISBN:

6 Cement Plants, Document DIA E-1996 [5] Brown-Boveri (BBC), Raw Material Quality Control, Document DIA E-1996 [6] Carp, H., Recent World - Wide Developments in the Area of Computer Control Systems for Cement Plants, IPA Scientific Session Bucureşti, 1982 [7] CEPROCIM S.A., Cement Industry Engineer Book, Editura Tehnică, Bucureşti, 1994 [8] Chamorro, Javier, Sistemos expertos basados en logica difusa, Cemento-Hormigon nr.705/1992, Espana [9] F.I.H., Critical Study of the Automatic Systems of Cement Production, S.C. Casial S.A. Deva, stage I, Research Contract no.195.1/1996 [10] F.I.H., Modern Solutions for Automatic Technical Production Systems, S.C. Casial S.A. Deva, stage II, Research Contract no.195.2/1996 [11] Precup, R. E., Fuzzy Controllers, Editura Orizonturi universitare, Timisoara, 1999 [12] Panoiu, M., Panoiu, C., Sora, I., Iordan, A., Study about the process control of an electric arc furnace using simulations based on an adaptive algorithm, WSEAS Transaction Science and Applications 5 (11), pp , 2008 ISSN: ISBN: