Mould design

Improving cooling inserts with evolutionary theory

Page: 3/3

Related Companies

Letting natural selection spawn the best solution

They are based on the idea of an optimisation using nature as the ideal example. Every life form evolves from its inception and strives towards an improved state with every generation. Therefore, a given optimisation problem can be improved over several generations of development. Based on a start population, the relevant parameters were randomly varied and combined. Afterwards, the capability of those recombined products of the population are evaluated via a target function. The best solution serves as the start population for the next generation, depending on the design of the algorithm. This sequence repeats until a suitable solution is found or a stop criterion is reached.

Cooling channel geometry evolves with each step

By transferring this method to the tempering channels of an injection mould, their exact course shall be evolutionarily varied and optimised. This leads to a channel course automatically generated by objective criteria. For example, one starting point is the variation of a single vertex of the tempering channel. This procedure can be extended to every other vertex and concludes in a different global tempering situation of the mould. However, criteria other than the positions of the vertices have to be included in the variation, for instance, the channel diameter or the number of vertices. The development of a proper target function therefore has special importance for finding and benchmarking a good solution. Every influencing variable that interacts with the tempering system has to be included in the parameters and mathematically described. Also, single factors have to be weighted for the target function. The precise weight will be calculated in this project phase in interplay with injection moulding simulations by first using simple geometry. The simulation will help to determine the effect of the variation of single factors. The quality of a specific course of the tempering channel can then be qualified by a single number through the target function. This method makes possible the ability to recognise good channel sections and reuse them in the following generations.

On to the next level

One of the hopes of the project is that the algorithms will be transferred to more complex geometries to analyse the effect for real tempering problems. The weighted factors have to be redefined during this stage, if necessary, so the algorithm will improve during the tests.

Using the algorithm in software will enable users to enhance mould design and save time in optimising the course of the tempering channels.