AI-driven cognition for advanced injection molding and industrial implementation

Abstract

Plastic injection molding has been an essential part of mass production in numerous industries for many years. However, this traditional production technique cannot provide sufficient efficiency and quality in today's competitive environment. With the growing emphasis on sustainability, the increasing use of recycled raw materials, rising turnover rates, and labor costs, an advanced and intelligent production process has become essential. This article proposes an AI-driven cognition, capable of operating independently of part geometry, raw material, and production equipment in the plastic injection molding. In pursuit of this objective, cavity pressure sensors are placed in the critical areas of the plastic injection mold. Using the data collected for each cycle, a reliable zone is identified to ensure the manufacture of high-quality parts. One of the key innovations of this study is establishing the relationship between fluctuations in the cavity pressure curve for both quality of the part and machine parameters. Based on this relationship, a CNN-based baseline knowledge learner has been developed to provide operators with actionable suggestions when the production process deviates from the reliable zone. The proposed method has been implemented with an accuracy of 98%. Following the development of the baseline knowledge, the proposed method was applied to two industrial applications. The task-oriented knowledge adaptation method was applied to these parts, which exhibit distinct characteristics regarding part shape, raw material, and quality criteria. The integration to the production site was achieved with an average accuracy of 95%.