Coupled Process Chain Modeling

In the pro­ject, the seman­tic inter­ope­ra­bi­li­ty of mathe­ma­ti­cal pro­cess models bet­ween indi­vi­du­al steps of the pro­cess chain is to be impro­ved. To this end, model exten­si­ons and a plat­form for data manage­ment as well as for the uni­form exe­cu­ti­on of the models will be deve­lo­ped and vali­da­ted in the indus­tri­al pro­duc­tion of alu­mi­ni­um and cop­per flat pro­ducts. A spe­cial focus is on the descrip­ti­on of inter­faces and data for­mats and the use of seman­tic annotations.

Logo of the IGF project

The basis lies in the depen­dence of the pro­per­ties of metal­lic and semi-finis­hed pro­ducts on the enti­re chain of pro­ces­ses pas­sed through during manu­fac­tu­re. The mecha­ni­cal pro­per­ties such as ten­si­le strength, hard­ness and tough­ness are influen­ced not only by the alloy com­po­si­ti­on but also by the degree of forming and the tem­pe­ra­tu­re-time pro­fi­le in the pre­hea­ting and heat tre­at­ment pro­ces­ses. The cur­rent pro­per­ties of the pro­duct also have an inver­se effect on the para­me­ters of the pro­cess being run through. The requi­red forces and maxi­mum forming degree per forming step are dic­ta­ted by the cur­rent mecha­ni­cal pro­per­ties. The heat trans­fer within the fur­nace is in turn depen­dent on the cur­rent pro­duct geometry.

Map­ping of the­se com­plex and mul­ti­di­men­sio­nal depen­den­ci­es bet­ween pro­cess para­me­ters and pro­duct pro­per­ties has so far been con­side­red in indus­tri­al prac­ti­ce only for indi­vi­du­al pro­ces­ses and not over the enti­re pro­cess chain. In indus­tri­al prac­ti­ce, pro­cess data from the pre­vious pro­ces­ses is usual­ly available when plan­ning and exe­cu­ting down­stream pro­ces­ses. Aggre­ga­tes for indi­vi­du­al pro­ces­ses are incre­asing­ly equip­ped with mathe­ma­ti­cal pro­cess models. Among other things, they ser­ve as soft sen­sors and deter­mi­ne values that are not acces­si­ble with mea­su­re­ment methods and pre­dict values that will occur in the future when a con­trol sce­na­rio is appli­ed. The know­ledge gai­ned can be used for model-pre­dic­ti­ve con­trol as well as con­ven­tio­nal con­trol, for batch plan­ning, when con­side­ring pro­per­ties like qua­li­ty and ener­gy savings. In forming pro­ces­ses, such models are used, for exam­p­le, for the design of ther­mo­me­cha­ni­cal pro­ces­ses, cal­cu­la­ti­on of the tem­pe­ra­tu­re via a rol­ling pro­cess or height-resol­ved grain size prediction.

Exem­pla­ry pro­cess chain

The exem­pla­ry pro­cess chain for strip mate­ri­al shown in the figu­re may con­tain a pro­cess model for each indi­vi­du­al aggre­ga­te, which has been adapt­ed from years of indus­tri­al use. With the use of such models or model­ling stra­te­gies, bes­i­des the con­side­ra­ti­on of the pro­duct pro­per­ties, an ana­ly­sis of the ener­gies and resour­ces used for each indi­vi­du­al pro­cess is pos­si­ble. The pro­blem, howe­ver, is that in indus­tri­al prac­ti­ce, once gene­ra­ted, data is usual­ly not available to pro­cess models of other plants, alt­hough the­se data is valuable as initi­al con­di­ti­ons and para­me­ters for models of pro­ces­ses down­stream in the pro­cess chain. Models that are sup­pli­ed as part of a plant and/or have been crea­ted or sup­pli­ed to address spe­ci­fic issues within a pro­duc­tion pro­cess are gene­ral­ly not desi­gned to be used in an over­ar­ching view across mul­ti­ple pro­ces­ses. The models some­ti­mes con­tain exten­si­ve infor­ma­ti­on about pro­ces­ses and, if they are updated or trai­ned manu­al­ly and/or auto­ma­ti­cal­ly with infor­ma­ti­on from pro­duc­tion, have been impro­ved and opti­mi­zed over years in some cases.

Tog­e­ther with the IBF and the IAT of the RWTH Aachen Uni­ver­si­ty, the IOB stri­ves to cou­ple the indi­vi­du­al pro­cess models via a sui­ta­ble plat­form for data manage­ment. The data exch­an­ge is inten­ded to impro­ve the accu­ra­cy of the pro­cess models and thus achie­ve ener­gy and resour­ce savings and qua­li­ty impro­ve­ments in the pro­ces­ses. Defi­ned inter­faces and data seman­ti­cs are inten­ded to sim­pli­fy the inte­gra­ti­on of hete­ro­ge­neous models in particular.