Intelligent control of ISBM process for recycled PET bottles

Abstract

Abstract. To manufacture plastic bottles with an increased ratio of rPET (recycled Polyethylene terephthalate), the ISBM (Injection Stretch Blow Moulding) process must be controlled to account for the variable mechanical and thermal properties. Calibration and optimization of the process have been successfully realized in past works but cannot be used for real-time applications. To address this, a gaussian process regression model of the free blowing step is created. It can calibrate itself using the pressure curve from a previous blowing to obtain near instantaneous predictions of key properties of the bottle. To create the model, the process’ characteristics are studied. Finite element simulations of the blowing where the properties follow a multivariate gaussian distribution are used to train the artificial intelligence. Then, an example is shown using the artificial intelligence predictions to optimize the thickness distribution of a bottle after blowing.