Data-driven calibration of temperature sensors for monitoring and control of injection molding machine

Monitoring the temperature in the plastic injection molding process plays a critical role in ensuring the quality and consistency of the final product. Thermocouples are widely used in injection molding machines to continuously monitor the process. However, accurate calibration of these sensors is essential for reliable and precise data acquisition. Traditional calibration methods often rely on manual procedures, which can be time-consuming, expensive, and require the shutdown for a prolonged time. In this paper, we present a data-driven approach for calibrating temperature sensors inside an injection molding machine without dismantling all the sensors that are embedded in the cavity. Our method exploits the repeatable relationship between the sensors to calibrate the sensors at the same time maintaining the traceability to the International System of Units. The proposed system provides better accuracy for sudden drift correction and reduces the number of physical calibrations required to maintain traceability to the reference standards. To evaluate the effectiveness of our approach, we calibrate a set of five sensors attached to different zones in an injection molding machine. The results demonstrate that our data-driven calibration method helps to reduce the number of physical calibrations while maintaining traceability to the standards. The method was evaluated for calibration and correction of sudden artificial drifts compared to existing data-driven methods such as multi-output Gaussian process and regression methods and found to provide better results in terms of accuracy and computational requirements.