A GA/gradient hybrid approach for injection moulding conditions optimisation

Injection moulding conditions such as melt temperature, mould temperature and injection time are important process parameters. Optimisation of these parameters involve complex patterns of local minima, which makes it very suited for Genetic Algorithm (GA). However, once a minimal region is identified during the search process, the GA method is not efficient, even sometimes impossible, in reaching its minimum. This is because GA is opportunistic not deterministic. The crossover and mutation operation may lead the search out of the identified minimal region. Gradient methods, on the other hand, are very efficient in this regard and can guarantee a local minimum, but not a global one. In this paper, a strategy of using a hybrid of both methods in injection moulding conditions optimisation is proposed, so as to exploit their respective advantages. The hybrid optimisation process is elaborated and a case study is conducted to test the effectiveness and efficiency of the strategy and its implementation algorithm. The optimisation results from the hybrid approach are compared with those from the GA method alone to demonstrate the improvement.