Modernizing Tooling Systems Through the Development of Hybrid Soft Sensors

Abstract Modern manufacturing increasingly relies on real-time data to optimize processes and ensure quality, particularly in tooling systems where direct measurement of critical variables is often challenging or impractical. This paper presents an approach to modernizing tooling systems, such as molds and dies, through the development of soft sensors. Specifically, this paper investigates the potential of developing hybrid models to predict hard-to-measure process variables or quality attributes using readily available machine data or easy-to-install sensors’ data. This approach is demonstrated in an injection molding case study, where a microstrain gauge installed on the exterior mold surface is employed to predict cavity pressure, a crucial parameter influencing product quality. This paper presents the results of a data-driven analysis and finite element modeling (FEM), as well as a methodology to integrate these results into a hybrid model to generate improved predictions of critical process variables from indirect measurements (i.e., soft sensors). Machine learning (ML) algorithms are trained on data from the microstrain gauge and cavity pressure sensors, providing predictions on quality metrics such as part weight, diameter, and thickness. The FEM model simulates the physical behavior of the mold under typical operating conditions, capturing the relationship between strain and pressure in detail. The hybrid modeling framework leverages the strengths of both approaches, combining physical insights with data-driven adaptability to enhance accuracy and robustness. Results demonstrate the challenges and limitations of both data-driven and physics-based models in such applications and highlight the potential of developing hybrid models using soft sensors to facilitate online quality monitoring and improve process control in manufacturing. By reducing reliance on offline inspections and manual quality checks, this approach offers a cost-effective solution for real-time quality monitoring, exemplifying the potential of soft sensors in modernizing tooling systems.