This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

PET bottles, produced at a staggering rate of 1 million per minute, significantly impact plastic waste, energy consumption, and sustainability. Manufacturers face the challenge of minimising material usage while ensuring containers meet performance demands such as top-load and burst resistance. The various process parameters in Stretch Blow Moulding (SBM) interact in complex ways that make it difficult to understand and control the process. This lack of control results in inefficiencies, leading to wastage and poor material distribution. Modern approaches prioritise sustainability by using simulation tools to optimize preform design and process conditions, reducing waste and reliance on trial-and-error methods. A series of forming simulations with different process parameters were conducted to produce virtual PET containers with varying material and modulus distributions. These virtual bottles were subsequently evaluated for empty top-load (ETL) performance, a necessary test to evaluate their ability to withstand stacking forces during filling and transport. A two-stage, bidirectional numerical model was developed to predict the structural performance of a bottle from process parameters, and to determine the optimal process parameters for a desired structural performance. The first stage of the model aimed to understand the influence of process parameters on the material distribution of a bottle. Inputs to the model were the processing parameters, and the target values were material distributions. Following hyperparameter optimization, a relationship was established. Through this intermediary step, there is enhanced interpretability into the manufacturing process. The second stage of the model involved developing the link between material distributions and the top-load performance of the bottle. Gaussian Process Regression (GPR), with the Radial Basis Function (RBF) kernel, was used to model this relationship. Through constrained global optimization, the required processing parameters were determined for a desired ETL performance. This model allows for the identification of optimal material distributions, thereby reducing wasted material, improving manufacturing efficiency, and providing new insights into the SBM process.