Exploiting data in smart factories: real-time state estimation and model improvement in metal forming mass production

Abstract

Modern production systems have numerous sensors that produce large amounts of data. This data can be exploited in many ways, from providing insight into the manufacturing process to facilitating automated decision making. These opportunities are still underexploited in the metal forming industry, due to the complexity of these processes. In this work, a probabilistic framework is proposed for simultaneous model improvement and state estimation in metal forming mass production. Recursive Bayesian estimation is used to simultaneously track the evolution of process state and to estimate the deviation between the physics-based model and the real process. A sheet bending mass production process is used to test the proposed framework. A metamodel of the process is built using proper orthogonal decomposition and radial basis function interpolation. The model is extended with a deviation model in order to account for the difference between model and real process. Particle filtering is used to track the state evolution and to estimate the deviation model parameters simultaneously. The approach is tested and analysed using a large number of simulations, based on pseudo-data obtained from a numerical sheet bending model.

A Hybrid Data-Based and Model-Based Approach to Process Monitoring and Control in Sheet Metal Forming

“…These opportunities are still quite under-exploited in automotive SMF, especially due to the complexity, scale, high tooling costs, and advanced process control involved. Even though literature clarifies the presence of several promising concepts related to sensor, tooling, and modelling [6] within metal forming, it has led to little industrial take-up [6,20], which depicts the difference between state-of-the-art and state-of-practice. Furthermore, solely collecting data in itself does not generate benefits.…”

“…Reference [6] points toward the need to consider the current state of the system while determining control response instead of utilizing a pre-planned control sequence built using offline process models. Furthermore, Reference [20] foresees the development of metal forming research toward a hybrid approach where physics-based models are combined with big data in manufacturing. On similar lines, Reference [19] points to the potential of providing significant improvements in the quality of a finished product through the combination of increased information about product properties and model-based control systems.…”

“…Additionally, Reference [29] envision the creation of systems that will combine Statistical Process Control (SPC), machine control, in-process control, and cycle-to-cycle control capabilities to significantly improve the part quality and consistency within the SMF process. In short, research [6,7,19,20,29] predicted the potential of benefiting from a combination of data-and-model-based approaches moving forward. In general, there is a lack of an overarching framework for monitoring and control of complex manufacturing processes, with only a few examples depicting application-specific approaches such as shown in References [6,38].…”

“…Furthermore, the control actions in such approaches are generally based on either offline models or online process models, which involve several approximations to be computationally feasible [19]. In both cases, the influence of product-process parameter correlations on the output product quality is overlooked by not explicitly modelling such complex relationships [20]. Thus, in the presence of changing production conditions, the overall effectiveness of control loops is affected, which leads to higher costs and process downtime.…”

Hybrid Modelling by Machine Learning Corrections of Analytical Model Predictions towards High-Fidelity Simulation Solutions

“…González et al (2019) [ 20 ] performed corrections for hyperelastic models based on data-driven machine learning, whereas Ibáñez et al (2018) [ 21 ] implemented a hybrid approach consisting of constitutive modelling and data-driven machine learning correction of plasticity models. In a manufacturing application example for metal forming production, Havinga et al (2020) [ 22 ] performed real-time predictions via a hybrid modelling approach that contains physics-based simulations those predictive deviations to the real process are eliminated via an additional corrective model. Overall, the specific employment of machine learning models alongside governing physics-based relationships allows for highly valid predictions within materials mechanics and its related fields.…”