Process control of a high-resolution infrared oven using evolutionary algorithms for processing sustainable lightweight materials

Procedia CIRP, vol. 139, pp. 1–6

Abstract

A high-resolution infrared emitter field has been developed for the energy-efficient and reliable processing of sustainable lightweight materials made from thermoplastic fiber-reinforced plastics with discontinuous thickness distributions. This consists of a large number of individually controllable micro-emitters whose thermal effect is recorded and evaluated using full-surface infrared thermography. The aim is to achieve a homogeneous temperature distribution in the semi-finished product in order to optimize the subsequent forming process. An AI-supported control model based on evolutionary algorithms was developed to determine the ideal power distribution. The CRISP-DM process (Cross Industry Standard Process for Data Mining) was used as part of a methodologically sound approach. The individual phases of this established method were used for structured modelling and targeted optimization of the AI system. The final algorithm follows a multi-stage procedure: In the first step, the radiator field is remodeled by analyzing which radiator influences which area of the semi-finished product. In the second step, the material and geometry-related influences of the semi-finished product are digitally mapped. On this basis, a forecast of the resulting temperature distribution can be created. Based on this, the evolutionary algorithm iteratively optimizes the emitter control in order to achieve a homogeneous temperature distribution in the semi-finished product. The integration of process-related data streams enables continuous model expansion and process control in a closed loop. The AI model developed was validated as part of a pilot production and critically evaluated in terms of its performance. In order to counteract the “black box” problem of many AI applications, central decision-making processes in the model were designed to be transparent and comprehensible. The use of explainable AI (XAI) not only increases trust in the system, but also strengthens social sustainability. The results show the potential of evolutionary algorithms for adaptive process control in the thermoforming of complex lightweight materials, particularly for applications in the automotive and aerospace industries, and provide a promising outlook for the large-scale industrial use of such systems for processing sustainable plastics.

Authors 5

1. Matthias Froning corresponding

   Institut für Textiltechnik Augsburg (Germany)

   Affiliation as printed

   Institut für Textiltechnik Augsburg gGmbH, Am Technologiezentrum 5, 86159 Augsburg, Germany

2. Maximilian Backenstos
   Affiliation as printed

   DatenBerg GmbH, Haid-und-Neu-Straße 7, 76131 Karlsruhe, Germany

3. Lisa Richter
   Affiliation as printed

   DatenBerg GmbH, Haid-und-Neu-Straße 7, 76131 Karlsruhe, Germany

4. Mesut Cetin

   Technische Hochschule Augsburg · Institut für Textiltechnik Augsburg (Germany)

   Affiliation as printed

   Institut für Textiltechnik Augsburg gGmbH, Am Technologiezentrum 5, 86159 Augsburg, Germany

   Technische Hochschule Augsburg, An der Hochschule 1, 86161 Augsburg, Germany

5. Thomas Gries Aachen Lehrstuhl für Textilmaschinenbau

   RWTH Aachen University

   Affiliation as printed

   Lehrstuhl für Textilmaschinenbau, Otto -Blumenthal-Straße 1, 52074 Aachen, Germany