Intelligent Temperature Control of a Stretch Blow Molding Machine Using Deep Reinforcement Learning

Abstract

Stretch blow molding serves as the primary technique employed in the production of polyethylene terephthalate (PET) bottles. Typically, a stretch blow molding machine consists of various components, including a preform infeed system, transfer system, heating system, molding system, bottle discharge system, etc. Of particular significance is the temperature control within the heating system, which significantly influences the quality of PET bottles, especially when confronted with environmental temperature changes between morning and evening during certain seasons. The on-site operators of the stretch blow molding machine often need to adjust the infrared heating lamps in the heating system several times. The adjustment process heavily relies on the personnel’s experience, causing a production challenge for bottle manufacturers. Therefore, this paper takes the heating system of the stretch blow molding machine as the object and uses the deep reinforcement learning method to develop an intelligent approach for adjusting temperature control parameters. The proposed approach aims to address issues such as the interference of environmental temperature changes and the aging variation of infrared heating lamps. Experimental results demonstrate that the proposed approach achieves automatic adjustment of temperature control parameters during the heating process, effectively mitigating the influence of environmental temperature changes and ensuring stable control of preform surface temperature within ±2 ℃ of the target temperature.