INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Improving scheduling methodologies in a Hot-Dip Galvanizing Line combining non-linear projectors and clustering
Andrs Sanz Garca, Francisco J. Martnez de Pisn Ascacibar, Rubn Lostado Lorza, Roberto Fernndez Martnez and Julio Fernndez Ceniceros

AbstractAn improving methodology for the scheduling coils of a Hot Dip Galvanizing Line (HDGL) is presented. This method uses a non-linear projector which has been selected from various techniques to generate a coil map from the most signicant parameters of the coils database: process variables, chemical composition of steel, measurements, etc. The created bidimensional map helps experts to decide which are the more tting groups showing the distances between all coils. After that, the expert can select with an end-user application one group and identify other coils that can complicate the scheduling purposes. Finally, the methodology uses hierarchical clustering to obtain a list of effective sequences of coils. A decrease of the number of shutdowns and irregular heat treatments failures can be obtained by using this scheduling method. Index TermsScheduling Methodology, Hot Dip Galvanizing Line, Hierarchical Clustering, Sammon Mapping, Kruskals nonmetric projector

Figure 1.

Basic scheme of a HDGL

HERE The constant effort to increase product quality [27] and reduce the expenses caused by failures in the manufacturing process is ongoing with Hot-Dip Galvanizing Lines, which are always trying to optimize their operative costs. A. Hot Dip Galvanizing Line This industrial process is composed by several stages. The initial material is the steel coils from the cold rolling with the required thickness. Broadly speaking, the process within a Hot-Dip Galvanizing Lines (HDGL) (Figure 1) can be described in the following
Manuscript received January 31, 2011: Revised version received March 8, 2011. This work was supported in part by La Rioja University through FPI fellowships and the Autonomous Government of La Rioja under Grant FOMENTA 2010/13. A. Sanz-Garcia is with EDMANS Group, Mechanical Engineering Department at La Rioja University, Logroo, La Rioja, 26004, Spain. (phone: +34941-299-273; fax: +34-941-299-794; andres.sanz@unirioja.es). F. J. Martinez-De-Pison-Ascacibar is with EDMANS Group, Mechanical Engineering Department at La Rioja University, Logroo, La Rioja, 26004, Spain. (fjmartin@unirioja.es). R. Lostado-Lorza is with EDMANS Group, Mechanical Engineering Department at La Rioja University, Logroo, La Rioja, 26004, Spain. (ruben.lostado@unirioja.es). R. Fernandez-Martinez is with EDMANS Group, Mechanical Engineering Department at La Rioja University, Logroo, La Rioja, 26004, Spain. (roberto.fernandez@unirioja.es). J. Fernandez-Ceniceros is with EDMANS Group, Mechanical Engineering Department at La Rioja University, Logroo, La Rioja, 26004, Spain. (julio.fernandezc@unirioja.es).

I. I NTRODUCTION

manner: The rst stage in the line consists of the formation of a continuous strip using steel coils that come from a mill process. Next, the strip passes through a preliminary cleaning section at the entrance of the annealing furnace, where it receives a heat treatment; the stage prior to its immersion in the liquid zinc bath. This treatment is essential in order to improve the properties of the strip and its coating. After the bath, the strip exits vertically, passing through blade-like currents of air which regulate the thickness of the cover. The temperature and cooling rates are controlled to obtain the desired mechanical properties for each steel type. Finally, the steel is run through a molten-zinc coating bath, followed by an air stream wipe that controls the thickness of the zinc nish (see Fig. 1). This description, with minor modications, is valid and applicable to all Hot Dip Galvanizing Lines in operation throughout the world. B. The Quality of the Galvanized Product The quality of the galvanized product, according to several authors [2], [19], can be divided into two fundamental aspects: the anti-corrosive characteristics and the properties of the steel. First, the anti-corrosive characteristics: steel comes marked by the thickness and uniformity from the zinc coating and basically depends on the supercial preparation of the metal bases, the control of the heat treatment, the bath composition and temperature, the air-blade control, the speed of the strip and the amount of time available for alloying at the specied temperature after the air-blade treatment. Secondly, the properties of the steel which fundamentally depend on the composition of the steel, the surface roughness, the process of smelting, the processes of rolling and nally,

Issue 5, Volume 5, 2011

890

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

the heat treatment that is applied to the strip before its passage by immersion in the liquid-zinc bath. By examining previously exposed coils, it can be deduced that the control of the heat treatment in a HDGL is fundamental for the process of coating and for the improvement of the properties of the steel. One of the key aspects concerning temperature control consists in assuring that the suitable thermal cycle is reached for each coil while considering the limitations of the HDGL furnace inertia. This is fundamental for the process of coating and for the improvement of the properties of the steel and can only be accomplished when the ow of the steel mass for unit of time is constant and the changes therein are carried out gradually. A fundamental coil sequence constraint is the need to limit changes in the mass ow rate from one coil to the next. But we must also examine other inconveniences. For example, coils within each group have to be sequenced based largely on decreasing sort by width or coils that have to be selected with similar steel properties and thermal requirements. Also, it must be borne in mind that defective coils can appear, or coils with characteristics that differ greatly from the others, and therefore, must be separated from the rest. For these reasons, it is crucial to order the coils that form the strip in such a way that the varying thickness, widths and types of steel do not differ too greatly [23], [31]. When this is accomplished, the temperatures generated by the furnace progressively t each individual coil. We see, therefore, that one of the most important tasks in the process in an HDGL is centered on obtaining sequences of coils that do not contain abrupt changes in the dimensions and types of steel in the consecutively treated coils. Logically, depending on the type of process, the demands of planning will be different. As has been mentioned above, one of the primary aims of the HDGL is to determine the best sequence of coils, after which it conforms to the changes in the new dimensions and types of steel. This is essential in assuring that the quality of the coating of each coil and the production are optimal. The present systems of planning in the HDGL divide the process into two general phases: The coils are grouped according to the type of steel (depending on its chemical composition) and its specications (specications of production, date of delivery of the product, specications of quality, etc.). Each group of coils is ordered such that the changes in width and thickness are minimalized. Normally, the process of coil scheduling begins by selecting those coils of greater width, and gradually moving towards those that are more narrow. Also, the HDGL looks for variances which are minimal, so as to create a smooth transition. II. P REVIOUS W ORK We review some of the present scheduling [4], [25] techniques in steel industry. There are several publications that deal with the problem of scheduling in an HDGL. The majority of the publications that address applications and techniques for scheduling in the steel industry are based on continuous
Issue 5, Volume 5, 2011

casting machines. The systems presently employed in these processes are generally based on either heuristic scheduling strategies or are implemented using knowledge software technology [23], [26]. Tang et al. [31] offer a review of the more commonly used methods and scheduling systems in the steel industry. In this work, different methods of optimization are described according to many planning systems (actually integrated in several industrial plants in the steel industry). In this article, the optimization techniques are grouped according to the following classication: Mathematical or heuristic models extracted directly from the analysis of the process. Methods [11] of articial intelligence (AI) divided as well into: a) Expert systems: based on the experience of the experts and the observed process restrictions. b) Random methods in search intelligence: genetic algorithms (GA), simulated annealing (SA) and tabu-search algorithms (TS). c) Constraints satisfaction methods (CSP) [32]. Human-machine coordination methods: consisting of the creation of scheduling by means of the iterative dialogue between human and machine. Multi-agents methods (A-teams methods): which rely on the cooperation of multiple models with different algorithms (each one of them with different constraints), so that the best solutions are found [10], [15]. For each kind of problem, multiple mathematical approaches and heuristic algorithms of optimization are selected [18], [29]. Other authors [9] focus the resolution towards problems of the type Traveling Salesman Problem (TSP) using the dimensions of each coil and other parameters. This methodology is based on the calculation of a triangular matrix of distances between all the coils to determine the best possible sequence. The most critical point of this consists in nding a trustworthy formulation that indicates the degree of existing similarity between two coils according to their physical and chemical properties, as well as restrictions in the process. Other works, however, are based on the use of graphical methods for scheduling. For example, Tamura et al. [28] looks for optimal ways in distribution maps where each coil is represented according to two coordinates: thickness and width. At the moment, these techniques are complemented with new techniques [6][17], [20]: Prize Colleting VRP models, IA, multi-agents, methods of man-machine iteration, or combinations of all of the above. Finally, other works study when scheduling problems are subject to unexpected events. In these cases, the steel plant needs a scheduling decisin that must be made in real time about the possible rescheduling or reordering of tasks. Recently years, new scheduling techniques, as known as hybrid dynamic scheduling systems, have been developed por solving this hard problem. Aytug et al. [1] used an evolutionary algorithm to modify an initial knowledge base, using results taken from the simulation of the manufacturing line. In that contribution they achieved that the system could react to unexpected events. Ikkai et al. [13] proposed a status selection method based on evolutionary algorithms to induce scheduling

891

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

knowledge from manufacturing lines.

Data Base (Coils):

1. Dimensional Parameters. 2. Chemical Composition. 3. Process Parameters. 4. ... Non-Linear Projection

Bidimensional Mapping Coils

Second, the present work makes use of multidimensional scaling (MDS) for exploring similarities or dissimilarities in order to generate a bidimensional map of the coils and then, a clustering technique is utilized to nd the local clusters. The clustering technique is able to reduce the amount of data items by grouping them, but projection methods are necessary to reduce the dimensionality of the data in the rst place. In addition, the projection can be used to show groups if a enough small output dimensionality (two or three dimensions) is chosen. A. Principal component analysis This mathematical procedure is used to show a set of data as a linear projection [21] on such a subspace of the initial data space. This conversion is dened in such a way that the rst principal component accounts for as much of the variability in the data as possible. This transformation is sensitive to the scaling of the initial data values and to outliers that produce large errors.
Table I I MPORTANCE OF COMPONENTS FOR PCA

2
Modify Zone B

Visual Selection Groups

Automatic Scheduling G roup B

Reselect Groups and Optimise Zone B. Recalculate Scheduling.

6
NOT

Standard deviation Proportion of Variance Cumulative Proportion

PC1 2.1394 0.2692 0.2692

PC2 1.6382 0.1579 0.4271

PC3 1.3419 0.1059 0.5330

Optimal Scheduling Zone B? YES

Thanks to the PCA and the knowledge based on expertise, it can be dened a set of uncorrelated and relevant variables (some chemical components of the steel). These are shown in Figure 3.

7
Saving Scheduling Zone B.
Figure 2. The diagram of the complete methodology
0.00 -80 -60 -40 -20 0 20

PC2

In this work consists of a several-step approach to the problem using multi-dimensional scaling and hierarchical clustering. In the following sections, another experience is described, results obtained are discussed and nal conclusions are drawn. III. M ETHODOLOGY The methodology that considers in this work, consist of an approach to the problem in seven steps. Figure 2 presents the complete and comprehensive ow chart explaining the proposed methodology to select the optimal coils scheduling sequence. The rst step consists in developing a data base with the n weighted parameters of each coil (considered important within the classication process). These parameters are the followings: Width and thickness of the coil. Objective temperature at the process. Chemical composition of the steel (C, Mn, Si, S, P, Al, Cu, Ni, Cr, Nb, V, Ti, B, N, Ceq).
Issue 5, Volume 5, 2011

629 312 628 329 271 275134 885180 543 300 298 308 580 331 295 892 894 273 307 129 1170 220 9631139 1140350 516 816 94 514 306507 9629803 781 328 46573 302 91 336 95 305 582 217 951 960941 294 315 276 272 301 508 289 324822311132 310 190 8841027 451 304 626 87 3 1062 203572 555 1063 107252971 249563 326824196 515120189 1112041015 11021217786 561 568 617407 621172 511593 717 77115737780 825 588 314186 891193 309108 105139386 334 512190 16011951041028 90011228640712980 585215 335338623 21113851 130105 566 564 843107 510 35843981012 680 48 318504317 1210 1209 274201 1211 340319 119693 818106 3312351174 686 728451171 7251229 724788988 723551 325214 761 10521175 1178978 1099921658 194 12361141 3471123 91461071 216 868105419 545 1028231203 7481055968 18681 1205 163207 738567718 759845 997796 737913918 592 361 44238360 3517471059 595213 357219540692 10201146 321185 127226549657 554 1171010416 789 3215872487 849323 125689550 34206665664 161184197AlTot 31722912714 806955975709 2995091202 337844893 320 558691 691149761013 173 591339826189 726990 170 103311183327 7311216967 13720353 1142142541 1381083989 679 351 862290115857698 6710388 840583 8421034 96 586841 212191 899928124850 984 1453418 635981 1253858945 143790708 225927450764368 479606 4801082 1081 208313 247123253713 210 716 8632231053140 864934411173 721 979 749531408 7451003715 7461016805 688 7561212800 423 839797950 1124 205 641139 903559466546379 9011210091225 500 110111014 942 662 231530477 35468122919 7250917 12521147 574 456 341 316 154148953687569 211778570 116324851 392 922437542 13121683 910804422 742719411 1206 2461075 124221378 90975159 11681086 1069381 192 23425544784 1208691144 1678543994 931720993 965 552 162 78188 182 10857551150 1233 861907703 646 158513 5781214233799751 920682695 56072655362 8831981035758 74411581001 1090483 2411224343 181 974 76 195 359562364 1134152 242896 166 178732596 44 589 472 584 167 10651032 734342 1169 923254485 9047436154811227 5906 93312231142 467986 387 385 118660 1183 255251 690 1188663 850 576779 3751079 1107656 9391156684 5751060 412 4248531176 5792791073654 1011 175 1213 926484 128532870383 710 2931044659996 10741218 935898 1153 263653 1019 1166752 239701118706987782 2621133 7305571251 929895 464556 1207539 9401177 932614 1152 261753 174 522 627 632 915260 908 12081000 944 916776 74040382 938650 897 49 1154478 1061 356854946417 1008565349 119616449 232865415 172 269 85 634 258448251219 697 14726956366 259959947348700 445678 977 925998 819 521270622 376486 820 285296633 1018 2248711221 787 17781332 187 747793538 369 954202 754 666 750777 10801025 961553 281222413 346 77577 1145393 ANCHO924860705699 612 1215454 4441148 344 36995571 736856 345 1231661 992 1091694 365 8731024943991 794 795 1197286C620 9486071186 1551220 8901155352693 1167 419 1043499266141 966 707 735458 767711992356241026 1042476 1238498 277278 12491163 771164 7973365 597 1119 257 1149 256 867 649 1113146648 701 667 855 1114537144 970 453 389 12561226 1120 157 82 587 1121 394 30459866 964 462 72 8028464 1126452 291004 3552671143 1185 221022 268881 446 443 447 441 770 470 621 N 936384 265506 930 1187 1110727409 985 5367391056711 999390 1200168876 440 949 875 86763 874581 611 619 610 1084937647 367 1108461 982 741 1165 473577 63 1011228 783 7291005 380 1097475 972 83 1031 S 639 10922521023420 1002 1222547 702 Ni12321230 363 287497 1070410 421 460 11179581151969 489 60927438651 973 608 911852 1129230 711130 652 84 631 Cu 1234 1076 983 1255 529 613 463 1531248 625109518159303 882 149 1164 1198 630 523 1066 245 1184 6181098 200 280535 594 1033Nb 888ESPENT 704 176 528 494 525 471 492 179 98333 199 165889400 1030 11311111 488 209 288 469 66 1029 97 1067 902 2341451250 228 1254 1017696 1162 11281122 282 54 403 243 1112 590 1049136 768 227 156 171 374 132 455 391 395 548 519 396491 150 517 373490 MATTEMPMED2 520 10501095 59 397496 1100 58244330 218 236 372264821151 533 398 495859 135 240 1077 802 Ti 905 9571237V 229 952685 56769297817 55 401 493 406765 762 1204 1038 801 505 169238 1093 100 99 766 1037827 1088 1125 1089 1136534886921078 118057404503 1179 60405133 61 3221246 1243 524 3991242 457 526 4021068 527 3711127 1036 1239 8871241 1040 Cr 1094 474 B 106488 1039 828 848 829 1109 1021 465 1096 1244 798 1041 1245 1247 1137 1240 1181 1182 760 Mn 283602 837 835 831 Si P 807 833600 6051189 834757 846 830 1191 603 808 1192 1190 1194 1045 847832 604 814 237 1193 599 838 1046 1048 812 811 1116 809 598 810 813

-0.05

-0.10

815

836 468 1047 73 601 671 126 501 482 10581006 292 1057 291 502 636 1160 774675 77374414 6371161 87810871115 431 432 433 877 673 1007 672 676 1104 792 791 638 668 1106 370 1105 677 644 1103 880669 428 427 429 430 879 434 435 436 674 426 645 643 642 1135 1134 1102 670 0.00

-0.10

-0.05 PC1

Figure 3.

Coils data set with 17 main components

As shown Figure 4, the rst three principal components (PCA1, PCA2, PCA3) doesnt explain most of the coils groups

892

-80

-60

-40

-20

20

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Finally, this MDS uses the power for Minkowski distance in the conguration space, so results can vary considerably from one computer to another. Basically, the rank correlation between calculated dissimilarities and plotted distances is maximized, allowing tied distances to not have identical plotted distances, only sequential ranks. In Figure 5, the bidimensional map generated represents the distances between all the coils to organize. C. Sammon non-linear projector The Sammon non-linear projector [24] is suited for use in exploratory data analysis. In the present work, it is used to visualize the equivalent Euclidean distances [3] of the coils to be treated within a 2 space. This algorithm employs non-linear transformations in order to map the original space, from the signicant n weighted parameters so that they can inuence the classication process, onto a low-dimensional visual space (2 or 3 ), attempting to preserve the Euclidean distances between coordinates of patterns (logically, other linear or non-linear multi-dimensional scaling methods (MDS) can be used) [8]. Let us denote the distance between two objects i and j in the space n by d , and the distance between their projections by ij dij ; Sammon mapping intends to minimize the error function known as Sammons stress: 1
i<j

Figure 4.

Tridimensional plot of the steel coils

in this work. Table I demonstrates that three principal axes of the PCA cover less than 55% of the cumulative proportion of variance, and therefore can not be used to project every of the coils. It can be seen that the linear projection doesnt work because , so we propose two alternative that does. B. Kruskals non-metric projector This is another form of non-metric MDS method [5]. First, a n-dimensional conguration to minimize the stress is chosen. Then, an iterative algorithm is used, which will converge most of times in few iterations (about 10), but due the complexity of the algorithm this procedure is slow for large datasets.

The next step consists in reducing the original data base with the n parameters of each coil to the most important p parameters. These chosen parameters are the following: Thickness of the coil. Objective temperature at the process.

Ei =

d ij

2 dij dij
i<j

d ij

(1)

1.5

1.0

MATSAM$points[,2] -0.5 0.0 0.5 1.0 1.5

MATISO$points[,2]

0.5

0.0

-1.0

-0.5

-0.5

0.0

0.5

1.0

1.5

-0.5

0.0 MATSAM$points[,1]

0.5

1.0

MATISO$points[,1]

Figure 5.

Bidimensional plot of Kruskals Non-metric MDS

Figure 6.

First bidimensional plot of Sammons mapping

Issue 5, Volume 5, 2011

893

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Chemical composition of the steel (C, Mn, Si, S, P, V, Ti, B, N). This new data set has better projection because there is only the most relevant variables for the rst two axes of principal component analysis performed previously. In Figure 7, the nal bidimensional plot of Sammons mapping with the new data is represented.

MATNORMPROY[,2]

This map of points allows one to visualize in an objective manner the equivalent distances between all the coils, and to classify them according to all the considered parameters properly weighted. Logically, the coils with very close parameters will appear tightly-packed, whereas those coils whose values vary greatly amongst themselves will appear more looselygrouped. Once all the coils in a two-dimensional graph have been projected, the integrated software system allows for iteration with a human expert for the selection of main groups of coils with similar characteristics (see Fig. 8). From the map of coils generated with software package, the new groups of coils with similar characteristics are dened visually with a contour of lines that can be adapted by the user. Those different coils that do not correspond to any dened group are detected (Fig. 9).

0.0 0.0

0.2

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

MATNORMPROY[,1]

Figure 7.

Second bidimensional plot of Sammons mapping

D. Grouping of data Once dened, the coils like points within a space n are projected by means of Sammon projector in a bidimensional space. This bidimensional generated map represents the equivalent Euclidean distances between all the coils to organize.

Figure 9.

Dened groups visually using a Sammons map

Figure 8.

Dened groups visually using a Sammons map

In this way, the degree of similarity between the coils and the obtaining of the standardized distance of its projections can be determined visually.
Issue 5, Volume 5, 2011

E. Cluster analisys of data When the main groups of coils are all selected, the coils of each group are locally grouped according to hierarchical clustering (see Fig. 10). This is one of the large number of methods that have been proposed for cluster analysis. Presently, cluster analyses appearing in the literatures are mainly divided in four groups: hierarchical methods, optimal partitioning, distribution mixtures and non-parametric estimation of local densities. A strictly hierarchical method depends on dissimilarity or similarity measures, providing a complete scheme of division. This algorithm [14], [7] uses a set of dissimilarities for the n objects being clustered. Initially, each object is assigned to its own cluster and then the algorithm proceeds iteratively, at each stage joining the two most similar clusters, continuing until there is just one single cluster. Figure 10 shows a classication tree or dendrogram where each node beneath the tree represents a coil from group B

894

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Figure 10. Dendrogram of Group B formed from hierarchical clustering with six clusters

Figure 12. in zone B

Local scheduling method applied starting of the greatest density

and each straight line represents the grouping of two clusters positioned by a Euclidean distance equal to the height of that straight line. The main advantage in using this type of graph is that the search for subgroups can be automated such that each individual coil does not exceed a determined distance. Partitioning into a given number of set of coils is found by merely cutting the dendogram at an suitable height. Also, abrupt jumps between clusters can be predicted.

chemical components of the steel, target temperature, etc. Once the local scheduling is made, it is necessary to verify the number of times that it surpasses a dened distance (number of gaps), as well as the maximum distances obtained between ordered coils. If the number of gaps is big or the distances between consecutive coils great, the human expert can take different courses of action until the results of each local scheduling are optimal: i.e. modify the contours that dene the groups, divide or fuse groups, choose another distance of clustering, etc.; or by means of techniques widely known: mathematical models, TSP, heuristic techniques, functions relating several variables, etc. One of the techniques that yields the greatest results consists in beginning the classication by starting at the point where the greatest density of coils exists (top of the mountain of densities) and continuing downward with those points that are less dense (see Figure 12). The process is analogous to descending from a mountain, starting at the peak and the ending at the foot. Finally, once all of the groups are ordered, those coils that could not be classied in any group can be eliminated, ordered separately, or included in some of the already existing groups. IV. R ESULTS A. Analysis of a previous historical data using the proposed methodology The rst interesting results arose when analyzing the projection of the 14 chemical components of the coils (C, Mn, Si, S, P, Al, Cu, Ni, Cr, Nb, V, Ti, B, N) for each one of the coils from previous scheduling and upon comparing them with the results obtained after the galvanizing process. The use of the created map allowed, with a single look and an objective form, the groups of existing coils to be determined as well as to detect those coils whose chemical compositions were completely different. Figure 13 presents the non-linear

Figure 11.

Different subclusters into group B

In Figure 11 are shown 6 subgroups whose Euclidean distance between coils does not exceed 0.15 and are distributed within a bidimensional space. The nal scheduling is made automatically for each one of the dened subclusters according to one or several of the parameters employed: width, thickness,
Issue 5, Volume 5, 2011

895

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Figure 13.

Sammon coils projection according to chemical composition

Figure 14. steel

Groups of coils according to the chemical composition of the

Sammon projection according to chemical composition and outliers coils that clearly appeared projected separately from the others. In addition, by extending the right part of the map, two main groups (cluster A and B) and three secondary ones (clusters C, D and E) are dened (see Fig. 14). This map can easily be used to classify new coils in one group or another. The method we present is expected to become a useful tool in HDGL applications as Figure 14 shows with 6 groups generated by clustering the coils. B. Decreasing of unexpected situations and process interrupts One of the most important problematic events, that worries to the engineers, happens when the industrial process of unexpected way is interrupted such as: breakage of the band, desaleaning, failures in the system, etc. However, the main advantage when using the proposed methodology is observed when we compare the distances between consecutive coils within the scheduling, since it can be used to help in the prediction of potential problems in the process. One of these possible causes that can be attributed to problems in an HDGL is produced when one of the welded coils that forms a strip consists of a steel with mechanical characteristics different from the others. Thanks to the use of the map, the expert can easily detect the presence of two consecutive coils with different steel compositions that are going to be processed one after another one. Also, the map can serve as a useful tool to visualize in each moment the type of steel in the coil that is being processed. The system can also be used to detect when a cross-section differs substantially from the coils that precede and follow it, or has a defective weld, etc. Another important problem is the breakage of the band. This event is one of the most dangerous contingencies by the high economic losses that produces. These fundamentally must to
Issue 5, Volume 5, 2011

the lost time in removing the defective band and feeding the line with a new band, as well as, by the lost material and the deterioration that takes place in the machinery. The proposed methodology can help to avoid this event by detecting all strips with mechanical characteristics very different from the others. C. Two cases of study with problematic events with chemical composition In Figure 15, are shown two cases of potential, problematic events is shown a case of potential, problematic events due to the inclusion of a coil with a steel different from the others within scheduling. Clearly, one can appreciate the utility in measuring and charting the chemical composition of the steel coils which are involved in the industrial process. D. Results of the Experience As has been mentioned, this methodology takes into account not only the chemical parameters of the steel coils, but also the dimensional parameters (thickness and width), target temperature, and other parameters of the process. In the tests, the following parameters of each coil were included: width and thickness, chemical composition of the steel and desired nal temperature; all of them properly weighted. Introducing the database composed of 2, 436 different coils, a nonlinear Sammon projection was made by obtaining a map similar to the one of Figure 9. The handling of the map of projected coils was very userfriendly due to the following reasons: The selection of the different groups from coils was extremely simple. It allowed the expert to detect those coils with parameters which differed greatly from the others. It helped to classify new coils easily within the process.

896

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

Once the groups of coils were selected, the scheduling for each one was generated semi-automatically according to hierarchical clustering. The results were such that each group was easily ordered according to the desired requirements. The number of gaps between coils was considerably reduced, as well as maximum distances between them. Even in the presence of a great number of gaps or distance between coils, this methodology proved highly effective in generating smaller groups to accommodate the gaps, and in adapting to the contours of greater distances.

Figure 15. A dangerous hypothetical event due to the presence of a coil from another steel family in scheduling

V. C ONCLUSIONS This article presents a successful experience in the use of an MDS to generate bidimensional maps from the multitude of parameters that factor into the process of the creation of scheduling within an HDGL. The use of these bidimensional maps and the dendrograms allows one to graphically group the coils with the least distance between one another, thus ensuring that their scheduling presents no signicant problems. An iterative process consisting of the following steps was proposed: projecting the coils according to the parameters selected in a bidimensional map, dening the groups of coils with the closest distances between them, selecting subclusters using hierarchical clustering and separately ordering them according to algorithms of already known scheduling. In this way, lists of safer coil combinations are obtained, abrupt jumps between consecutive coils are reduced and potential problems spotted before occurring. R EFERENCES
[1] H. Aytug, G.H. Koehler, J.L. Snowdon, Genetic learning of dynamic scheduling within a simulation enviroment,Comput. Oper. Res. 21(8), 1994, pp. 909925.

[2] V. L. Baum, J.M Stoll, LTV stell hot dip galvanizing line upgrade for exposed automotive products,Iron and Steel Engineer 71, 1994, pp. 32 37. [3] S.-H. Cha, Comprehensive Survey on Distance/Similarity Measures between Probability Density Functions. International Journal Of Mathematical Models And Methods In Applied Sciences, Issue 4, vol. 1, 2007, pp. 300307. [4] J. I.-Z. Chen, C.-C. Yu, Estimate to the Trajectory of Maneuvering Targets by Combining Sensor Scheduling with Energy Efcient in WSNs. International Journal Of Mathematical Models And Methods In Applied Sciences, Issue 2, vol. 3, 2009, pp. 7785. [5] T. Condamines, The noisy multidimensional scaling problem: an optimization approach. In CSCC 2000 Multiconference Proc., ISBN: 9608052-19-X, 2000, pp. 24112416. [6] P. Cowling, M. Johansson, Optimization in Steel Hot Rolling. Optimization in Industry. European Journal of Operational Research, 1391, 2002, pp. 230244. [7] R. Dandrade, U-Statistic Hierarchical Clustering. Psychometrika 4, 1978, pp. 5867. [8] M. Daszykowski, B. Walczak, D.L. Massart, Projection methods in chemistry. Chemometrics and intelligent laboratory systems 1345, 2003, pp. 116. [9] J. Dorn, M. Girsch, G. Skele, W. Slany, Comparison of iterative improvement techniques for schedule optimization. European Journal of Operational Research 25, 1996, pp. 349361. [10] D. Frey, J. Nimis, H. Wrn, P. Lockemann, Benchmarking and robust multi-agent-based production planning and control. Engineering Applications of Articial Intelligence 16, 2003, pp. 307320 [11] B. Grabot, Articial intelligence and soft computing for planning and scheduling: how to efciently solve more realistic problems. Engineering Applications of Articial Intelligence 14, 2001, pp. 265267. [12] I. Harjunkoski, I.E. Grossmann, A Decomposition Approach for the Scheduling of a Steel Plant Production, Computer & Chemical Engineering 25, 2001, pp. 6471660. [13] Y. Ikkai, M. Inoue, T. Ohkawa, N. Komoda, A learning method of scheduling knowledge bu genetic algorithms, In: IEEE Symp. Emerging Technol. Factory Automation, 1995, pp. 641648. [14] S. Johnson, Hierarchical clustering schemes. Psychometrika 2, 1967, pp. 241254 [15] A. Karageorgos, N. Mehandjiev, G. Weichhart, A. Hmmerle, Agentbased optimisation of logistics and production planning. Engineering Applications of Articial Intelligence 16, 2003, pp. 335348 [16] B. Lally, L. Biegler, H. Henein, A model for sequencing a continuous casting operation to minimize costs. Iron and Steelmaker 14, 1987, pp. 63-70. [17] L. Lopez, M. W. Carter, M. Gendreau, The hot strip mill production scheduling problem: A tabu search approach. European Journal of Operational Research 106, 1998, pp. 317335. [18] S. Moon, A. N. Hrymak, Scheduling of the batch annealing process, deterministic case. Computer & Chemical Engineering 23, 1999, pp. 11931208. [19] J.B. Ordieres Mer, A. Gonzlez Marcos, J. A. Gonzlez, V. Lobato Rubio, Estimation of mechanical properties of steel strip in hot dip galvanising lines. Ironmaking and Steelmaking 1, 2004, pp. 4350. [20] H. Park, Y. Hong, S. Y. Chang, An efcient scheduling algorithm for the hot coil making in the steel mini-mill. Production planning & control 13, 2002, pp. 298306. [21] N. Pessel, J.-F. Balmat, Principal Component Analysis for Greenhouse Modelling. WSEAS Transactions on Systems Issue 1, vol. 7, 2008, pp. 2430. [22] P. Petr, J. Krupka, S. R. Provaznkova, Multidimensional Modeling of Cohesion Regions. International Journal of Mathematical Models and Methods in applied techiques Issue 1, vol. 5, 2011, pp. 298306. [23] D. J. Renn, J. L. Stott, J. F. Vasko, Penalty-based sequencing strategy implemented within a knowledge-based system. Journal of the Operational Society. Journal of the Operational Society 50, 1999, pp. 205 210 [24] J. W. Sammon, A nonlinear mapping for data structure analysis. IEEE Transactions on Computers 18, 1969, pp. 401409. [25] B. Scholtz-Reiter, J. Heger, M. Ltjen, A. Schzweizer, A MILP for installation scheduling of offshore wind farms. International Journal Of Mathematical Models And Methods In Applied Sciences, Issue 2, vol. 5, 2011, pp. 371378. [26] M. Soo Suh, Y. Jae Lee, Y. Kwan, A two-level hierarchical approach for raw material scheduling in steelworks. Engineering Applications of Articial Intelligence 10, 1997, pp. 503515.

Issue 5, Volume 5, 2011

897

INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES

[27] J.. Srisertpol, S. Tantrairatn, P. Tragrunwong, V. Khomphis, Estimation of the Mathematical Model of the Reheating Furnace Walking Hearth Type in Heating Curve Up Process International Journal Of Mathematical Models And Methods In Applied Sciences, Issue 1, vol. 5, 2011, pp. 167174. [28] R. Tamura, M. Nagai, Y. Nakagawa, T. Tanizaki, H. Nakajina, Synchronized scheduling method in manufacturing steel sheets. Intelligent Transactions Operational Research 3, 1998, 189199. [29] L. X. Tang, J. Y. Liu, A. Y. Rong, Z. Yang, A mathematical programming model for scheduling steelmaking-continuous casting production. European Journal of Operational Research 120, 2000, pp. 423435. [30] L. X. Tang, J. Y. Liu, A. Y. Rong, Z. Yang, A multiple traveling salesman problem for the hot rolling scheduling in Baoshan Iron and Steel Complex. European Journal of Operational Research 124, 2000, pp. 267282. [31] L. Tang, J. Liu, A. Rong, Z. Yang, A review of planning and scheduling systems and methods for integrated steel production. European Journal of Operational Research 133, 2001, pp. 120. [32] H. Wai Chun, Scheduling as a multi-dimensional placement problem. Engineering Applications of Articial Intelligence 9, 1996, pp. 261-273.

Julio Fernndez Ceniceros , received the M.Sc. degree in Industrial Engineering in 2009 and the Master Degree in Project Management from the University of La Rioja (Spain) in 2010. He is working at the University of La Rioja as a fellowship and his current research is applied numerical simulations (Finite Element Method) and Data Mining techniques in steel and concrete structures. His interests include the behavior of bolted connections, FEM simulations, failures modes, plasticity and damage.

Andrs Sanz Garca , received his B.S degree in Industrial Engineering in 1999 and M.S degree in Industrial Engineering from Universidad de La Rioja in 2002. Currently he is engaged on his PhD in the area of industrial optimization through soft computing techniques. He is registered Professional Engineer until 2011 in La Rioja, Spain. Recently, he is a professor and researcher of the Universidad de La Rioja in the Mechanical Department with EDMANS research group. He has published several papers in conferences. His research area includes but not limited to the use of nite element analysis to solve engineering problems, materials science, fracture mechanics, and failure analysis of engineering materials.

Francisco J. Martnez de Pisn Ascacbar , PhD - graduated in Industrial Engineering from Universidad de La Rioja in 1999. In 2003 he completed a PhD degree at the same university in the area of industrial optimization through data mining techniques. He is teacher and researcher of the Universidad de La Rioja in the Mechanical Department. His main research interests are data mining; soft computing; pattern recognition; articial intelligence and industrial optimization. He has published many papers in journals and is author of several books and industrial patents. He has been the main research of several national projects and has participated in other nationals and European data mining Projects. Currently supervises several MSc and PhD students in the area of data mining and soft computing.

Rubn Lostado Lorza , received the M.Sc. degree in engineering in 2003 and the Ph.D. degree in engineering from the University of La Rioja, Logroo, Spain in 2010. From 2003 to 2005 he was working on numerical simulation (Finite element method) at the department of Wind Energy in the company M-torres industrial designs. From 2005 to 2007 he was working on numerical simulation (Finite element method) in the Automotive Technological Center of Navarra (CITEAN). He is currently working at the University of La Rioja as professor of engineering projects and environmental technology. His current research is the combined modeling of industrial processes and products using the Finite Element Method and Data Mining.

Roberto Fernndez Martnez , MSc - graduated in Industrial Engineering from Universidad de La Rioja in 2006. He is working in his PhD degree at the same university in the area of industrial optimization through data mining techniques. He is researcher of the Universidad de La Rioja in the Mechanical Department. His main research interests are data mining; soft computing; pattern recognition; articial intelligence; industrial optimization and feature selection. He has been researcher of several data mining Projects. Currently is nishing his thesis in the area of data mining and soft computing.

Issue 5, Volume 5, 2011

898