Data processing technique and its applications in sheet metal forming processes

Abstract

In recent years, Finite Element Analysis (FEA) has been widely used by the metal forming industry in conjunction with traditional practical techniques, resulting in an accumulation of vast volumes of FEA results. With the increasing trend for the application of cloud computing services to FE software, the traditionally independent manufacturing sectors are being transformed into cloud connected communication networks through horizontal and vertical integration solutions. Industrial Research is shifting towards an eco-friendlier approach through exchanging, sharing and utilising of these data. Advancements in these newly developed technologies have brought new challenges to the sheet metal forming industry, leading to the emergence of a unique research field of data processing technologies for metal forming processes.

One of the most significant issues in the field is the extraction of insightful information from information-absent FEA data. In this work, the focus is on a dataset comprising of vast volumes of experimentally verified FEA data from various part geometries. Several data classification algorithms were developed using physical-based models and machine learning models, respectively, enabling a digital characterisation of finite elements from sheet metal forming process simulations in the absence of spatial or visual data. Machine learning models were developed to recognise the geometric information by over 90 percent accuracy. Based on the understanding of thermo-mechanical history of sheet metal forming processes, including stress state, such as the Lode parameter and stress triaxiality, several physical-based algorithms were then developed which were rigorous in recognising the flange zone, flat bottom and side wall with an excellent accuracy.

In addition, research in metal forming processes has traditionally followed the convention of validating hypotheses with repeatable experimental trials. However, the reliable characterisation of boundary conditions remains challenging, which can be partly explained by the fact that the real boundary conditions in metal forming processes are not well understood, especially for coefficient of friction (COF) and interfacial heat transfer coefficient (IHTC). A systematic data visualisation technique was developed to generate a tribo-map and an IHTC-map, characterising the thermo-mechanical history of a sheet metal process, including relative sliding distance, contact pressure and temperature difference between the workpiece and tool, which enables an insight into the evolutionary boundary conditions at different workpiece-tool interfaces. Such a data-driven approach is also adopted in this work to shape the experimental strategies by introducing variable contact/sliding conditions, and the friction of coefficient and interfacial heat transfer coefficient obtained at these variable testing conditions were found to have excellent match with numerical models.