GENETIC PROGRAMMING AND CAE NEURAL NETWORKS APPROACH FOR PREDICTION OF THE BENDING CAPABILITY OF ZnTiCu SHEETS

ZnTiCu (zinc-titan) sheets with approximately0,1%Cuand0,1%Ticontentareverywidelyusedintheconstruction industry for roof covering, gutters, drainpipes, facing linings, connections, window shelves, dec-orative elements on roofs, art products, etc. Data on theproduction technology of zinc-titan alloy sheet and onits forming properties are very scarce and unreliable.Therefore they must be checked for each individualtechnological step and the conditions under which themetal sheet is formed. Their forming properties are in-fluenced by many parameters, e.g. chemical composi-tion, technological parameters of rolling, etc. Due tolarge number of influential parameters the desired me-chanical properties of the metal sheet (e.g. bending ca-pability) are difficult to monitor and to keep within ac-ceptable technological limits. Rolling mills usually col-lect data on an individual batch (e.g. alloy composition,conditions in which the sheet metal has been rolled etc.)butinmostcasesthegeneralapproachassuringachieve-ment of the desired forming properties based on the in-fluential parameters of metal sheet production cannot betraced. Often it is not known which parameters are ofimportance.Insuchcaseslinearregressionsmethodsarenot efficient since the abundance of input parametersand their mutual influences make the determination ofan adequately precise model impossible 1 .In the present work two different approaches basedon experimental data on ZnCuTi alloy composition andon technological parameters of hot and cold rolling havebeen used to predict the metal sheet bending capability.The first one is the GP which belongs to the class of themethods of evolutionary computation 2-7 , and the sec-ond one is the CAE neural network, which has been suc-cessfully applied for solving many engineering prob-lems e.g. 8-16 .