Optimization in polymer processing
An introduction to optimization in polymer processing based on extrusion. Described methods are universal and can be applied in other polymer processes. We start from definition than optimization methods, how we acquire input and output date and for the end input and output data in extrusion.
Every task to do we can present as an answer. Optimization would be looking for the best (extremum) in the possible answers with assumed criteria, defined target function (e.g. max. output, min. power consumption).
For small area statistical methods like regression analysis when full area is scanned is applied. For large area artificial intelligence methods have to be applied, e.g. neural networks, genetic algorithms and hybrid methods (connect more than one). Extrusion optimization is complex. There are many optimization criteria, they can be conflicting like max. output and min. power consumption and many input data (geometrical, process parameters).
We have two ways of collecting the output data for optimization. Experiments and simulations. First can not be applied in some situation like changes in screw design. Second does not have this limitation and it can be effective tool for acquiring the data.
The characteristic process for optimization is the constant development which you can see on the picture. Output data from experiment or simulation as optimized can be applied as input data for other research.
This is the introduction. In the near future DOE (design of experiments) will be described as a one of the optimization technique.
From one site we have simulations and experiments one and other cost, experiments we know why, and simulations (e.g. REX, Moldflow, Cadmould 3D-F, etc.) software availability, not cheap. On the other hand, optimization the same software availability (e.g. Statistica). The best would be both simulation and optimization integration. What is more know how. Does it limit anything? Investment which will return. This is the way of doing business. Would you like to know more please let me know in mail or in comment. I will try to add what is needed.
Every task to do we can present as an answer. Optimization would be looking for the best (extremum) in the possible answers with assumed criteria, defined target function (e.g. max. output, min. power consumption).
For small area statistical methods like regression analysis when full area is scanned is applied. For large area artificial intelligence methods have to be applied, e.g. neural networks, genetic algorithms and hybrid methods (connect more than one). Extrusion optimization is complex. There are many optimization criteria, they can be conflicting like max. output and min. power consumption and many input data (geometrical, process parameters).
We have two ways of collecting the output data for optimization. Experiments and simulations. First can not be applied in some situation like changes in screw design. Second does not have this limitation and it can be effective tool for acquiring the data.
The characteristic process for optimization is the constant development which you can see on the picture. Output data from experiment or simulation as optimized can be applied as input data for other research.
This is the introduction. In the near future DOE (design of experiments) will be described as a one of the optimization technique.
From one site we have simulations and experiments one and other cost, experiments we know why, and simulations (e.g. REX, Moldflow, Cadmould 3D-F, etc.) software availability, not cheap. On the other hand, optimization the same software availability (e.g. Statistica). The best would be both simulation and optimization integration. What is more know how. Does it limit anything? Investment which will return. This is the way of doing business. Would you like to know more please let me know in mail or in comment. I will try to add what is needed.
Comments