The future of injection molding 

The test series for the Dar­WIN research project have been run­ning since the end of 2020. The aim of the AI project is to learn detailed behav­ior mod­els of injec­tion mold­ing machines on high-fre­quen­cy machine data. To do this, machines from dif­fer­ent man­u­fac­tur­ers that pro­duce sim­i­lar parts over time are used. In this way, the behav­ior mod­els learned on one machine should also be trans­fer­able to oth­er machines with­out hav­ing to learn the mod­els again com­plete­ly for each machine. The behav­ior mod­els sug­gest opti­mized process para­me­ters for the next shot in order to pro­duce with­out scrap at the min­i­mum pos­si­ble cycle time. 

Comparing injection molding machines and processes — independent of manufacturer 

The experts from SKZ and plus10 are con­duct­ing appli­ca­tion-ori­ent­ed research into the lat­est machine learn­ing mod­els for describ­ing the behav­ior of cycli­cal man­u­fac­tur­ing process­es using injec­tion mold­ing as an exam­ple. The focus is on online capa­bil­i­ty, i.e. the for­ma­tion and exten­sion of a mod­el while the process is run­ning. In addi­tion, the inves­ti­ga­tion of the trans­fer­abil­i­ty of pre-trained machine learn­ing mod­els from one machine to sim­i­lar, non-iden­ti­cal machines also plays a cen­tral role.

An “evo­lu­tion learn­er” from the com­pa­ny plus10 gen­er­ates opti­miza­tion sug­ges­tions based on behav­ioral com­par­i­son with all iden­ti­cal or sim­i­lar machines involved. SKZ pro­vides a wide range of machines from the man­u­fac­tur­ers ARBURG, ENGEL, Krauss­Maf­fei, Sum­it­o­mo (SHI) Demag and WITTMANN BATTENFELD for the test series.

plus10’s exper­tise in intel­li­gent data pro­cess­ing and auto­mat­ed pro­duc­tion opti­miza­tion by means of con­tin­u­ous­ly learn­ing mod­els will flow into the project. The injec­tion mold­ing spe­cial­ists eval­u­ate the trans­ferred opti­miza­tion pro­pos­als and check the com­po­nent qual­i­ty in the test laboratory. 

The “Dar­WIN” research project (BMBF fund­ing code 01IS20066) is fund­ed by the Ger­man Fed­er­al Min­istry of Edu­ca­tion and Research (BMBF) and is expect­ed to end in Novem­ber 2021. Pub­li­ca­tion of the final results is planned for the end of the year.