NEW SOFTWARE TOOL DPP BASED ON NEURAL NETWORKS
13 August 2007 - Bayer MaterialScience AG

"Designers and processors of our materials can use the patented software to achieve considerable savings, because it shortens the development time needed for molded parts, reduces the number of prototypes and cuts the number of injection molding trials," explains Olaf Zöllner, head of the Injection Molding Technologies Group in the Polycarbonates Business Unit. The program's strength is its flexibility and speed. The user can, for example, vary the processing parameters in the computer any time and see immediately how the properties of the molded part change as a result. "The engineer enters the specifications for the part into the program, and, on the basis of this, the software comes up with a proposal for the material to be used and calculates the optimum design and processing conditions," says Zöllner.

DPP is divided into several modules. With the design module, the mechanical characteristics of the thermoplastic under consideration can be determined under service conditions. It supplies stress-strain curves and secant modules, gives the loading limits and plots the influence of temperature. "For the first time, it is also possible to show the directional dependence of the mechanical properties over the entire part," says Zöllner. With the processing module, it is possible to establish how a specific design can best be injection molded. It calculates key processing parameters such as the filling pressure at various melt temperatures, the metering time as a function of screw speed and the level of shrinkage during processing. The rheological module provides information on the melt viscosity as a function of shear rate and melt temperature. The mechanical module refines the data obtained in the design module and analyzes their dependency on processing and the geometry of the part. For wall thicknesses between 1.5 and 4 mm and for the selected process conditions, it performs virtual tensile tests and supplies the results in the form of a stress-strain diagram. In the fifth module, the virtual lab, the DPP user can use predefined data to search for a suitable thermoplastic for an intended application.

The new software functions on the basis of a neural network. In simplified form, this imitates the processing of data by neurons in the human brain. The artificial neurons receive information from their neighbors, evaluate it and then decide which data to pass on to where. If a piece of information is entered into a network of this kind, it is processed into a result that can be read out on the "output neurons". So that this functions, the neurons and their links to each other must be programmed beforehand, namely with "specimen examples", for which the input data and results are known. With DPP, this happens all the time, among other things with a wealth of material data and processing data.

DPP is installed on a central server at Bayer MaterialScience. Further information on the conditions of use and online instructions on how to use it can be found at http://www.plastics.bayer.com/plastics/emea/de/dpp/index.jsp. After registering and being issued with a password, customers can interactively access the software from their PC with a locally installed Web client via a direct Internet connection.

http://www.bayermaterialscience.com

More News:
  • For August 2007
  • From Bayer MaterialScience AG
  • For Engineering Plastics