Methods for design and modelling of (multi-material) joints

Introduction

Selecting the right technique to join two materials together is a complex task, which has severe consequences not only for the design and the final product quality but also for the production process. Several joining technologies exist, such as bolting, riveting, gluing, spotwelding, clinch-bonding, rivet-bonding, weld-bonding etc. To make the right selection you need knowledge about all joining technologies but this covers so many aspects that this is infeasible. Most experts are therefore only specialized in one or a couple of these joining techniques.

Product designers or engineers, on the other hand, know a little about all joining techniques but lack further knowledge to make a proper selection especially for very specific or complex cases. Within the Jointtool project, an Online Knowledge Platform (OKP) was developed to bridge this gap.  The OKP provides general information about the different joining technologies, together with a selection tool which acts as a first quick filter to narrow the search of possible joining technologies. This information can be used by product designers or engineers to make a rough selection of possible joining technologies and next contact the right expert to make a final selection. During this webinar we will show which information is available on the platform and how the selection guide can be used to get an overview of interesting joining technologies.

Virtual Testing is a model-based approach to support joint design on a system level. The computational cost of detailed finite element models of joints (e.g. bolts or adhesive layers) might render them unfeasible to include in a model of the complete product under design. Concept models (simplified models specifically chosen for the application) for joints do not suffer from this computational limitation, making them feasible for system level simulations. Traditionally, the use of concept models is held back by the difficulty to obtain accurate parameters. Concept models usually have input parameters that do not have a direct link with physical parameters of the bolt or the adhesive (e.g. the parameters contain a generic stiffness instead of a Young’s modulus).

Virtual Testing can be used when validated detailed models are available (e.g. from a previous generation product), as parameters are extracted by performing virtual tests on them. This has the main advantage that no dedicated physical tests are required to obtain the joint concept model parameters. Additionally, sensitivity analyses can easily be performed on the models to identify the most impactful parameters and deeper insight into the physics of the joint can be gained through observation of quantities like stresses that are difficult to measure physically.

After performing the Virtual Testing method, system models of the complete product, including a conceptual but accurate description of the joints, can be calculated in an efficient way. The method is validated on bolted joints and adhesive joints. The methodology has been set up based on finite element models, but it is not restricted to it.

The accuracy of a Finite Element predictive model for an assembled system is highly dependent on the accuracy of the sub-models representing the structural joints, such as bolted connections and adhesive bondings. Within the JOINTTOOL project, the Virtual Sensing approach is proposed to improve the experimental methods used for the identification of these structural joint models. With Virtual sensing it is possible to measure at unmeasured locations of the testrig, enabling an augmented monitoring during the test and a more accurate estimate of the outputs. The methodology requires the Finite Element model of the testrig and few non intrusive measurements. These information are coupled in an online fashion, where timestep by timestep the status of the testrig is predicted with the model and then corrected with the measurements. In other works this approach is also referred as model-based testing, hybrid testing or augmented modeling.

In particular two virtual sensing tools have been developed for the estimation of adhesive parameters. The first one applies on vibration based tests, where the bonded structure is excited with the impact of an instrumented hammer and measured with few accelerometers. The identification of the adhesive Youngs’ modulus is based on the minimization of the error between real and virtual accelerometers. The identified parameter allows to reduce the error on the first modal frequency from 47% (Youngs’ modulus directly from adhesive datasheet) to 2% (identified with Virtual Sensing). The second virtual sensing tool applies on quasi-static lap shear tests (e.g. ASTM D1002, ISO 4587), where a single lap joint sample is gradually loaded in shear and the force and displacement at its ends are measured. The Youngs’ modulus extracted on the simple sample is used as an input parameter for the model of a more complex geometry where the total stiffness error is reduced from 110% (datasheet) to 22% (identified). Furthermore the tool enables the full field monitoring of the testrig, tracking entities that could not be measured in a traditional test setup such as the deformation of the substrates, the rotation of the joint and the peel and shear local loads.

In the endeavor to lightweight structural design to reduce pollution, the use of fiber reinforces plastics is indispensable. These materials demand for other joining techniques. One of these joining techniques is structural adhesives. Although the use of structural adhesives is not new, the dimensioning of the joint is still a cumbersome task. Simulation tools are available but they are either focusing on the overall structural behaviour or the joint itself in detail. Certainly when lifetime assessment is important, it is even more difficult to determine the joint strength. This research will give an overview on how adhesive joint dimensioning influences the joint strength and how we can handle lifetime assessment by incorporating ageing effects.

Finite element (FE) models are often employed to predict the dynamic behavior of jointed structures due to their high accuracy and direct link with physical parameters. However, full-scale FE models would result in large model sizes and prohibitively long simulation times for real-life applications. Instead of turning to conventional concept models for the joints, which are not available for each application, model order reduction (MOR) techniques can be applied to significantly reduce the computational costs of a FE joint model while maintaining their high accuracy and keeping the link with the physical parameters. These models can be further extended with parametric model order reduction (pMOR) strategies to enable one reduced model to represent the joint for a range of its parameters.

In this work, pMOR based concept models are developed for two popular types of joints, the adhesive joint and bolted joint. They contain local nonlinearities, such as the viscoelasticity of adhesive joints and contact nonlinearity of bolted joints, that influence the dynamic properties of structures and can be time consuming to simulate. By employing pROM concept models for these joints, computational costs can be reduced by up to 70% while still maintaining a high accuracy when compared to FE joint models.