Multi response optimization in vulcanization process using backpropagation neural network-genetic algorithm method for reducing quality loss cost

Rubber material has been widely used in footwear sole industry due to the effect of better mechanical properties. Vulcanization process is the main process of rubber sole manufacturing that affects the rubber sole quality. Tensile strength and elongation at break are some responses used to evaluate the performance of vulcanization process. The quality characteristics of these responses are the larger the better. Experiment was conducted to identify the combination of process parameters in vulcanization process. Three important parameters, namely: mold temperature, mold pressure, and holding time, were used as the factors. Each factor was set at three different levels. Therefore, 3 x 3 x 3 full factorial was used as design of this experiment. It later was replicated three times along with randomization. The optimization was conducted by using backpropagation neural network and genetic algorithm. The architecture of developed network indicated 3 (three) neurons on input layer, 16 neurons on hidden layer, and 2 (two) neurons on output layer. The activation functions of hidden layer, output layer, and network training were tansig, purelin, and trainlm, respectively. The maximum tensile strength and elongation at break could be obtained by using mold temperature, mold pressure, and holding time of 145°C, 84 bar, and 4 min, respectively. The total reduction of quality lost cost of the process was Rp 238.55 or 27.30% of quality loss cost before optimization.